Estimating treatment need by analyzing Google search queries shows promise
No approaches currently exist to capture the number of individuals who are looking for addiction treatment in a given area. Internet data could offer a low-cost and near real-time way to estimate help-seeking for substance use disorders, so that treatment resources could be made available more efficiently. This study examined characteristics and trends in Google searches focused on help-seeking for a variety of substance use problems conducted over a 10-year period.
Demand for addiction treatment, also known as help-seeking, is currently measured in the United States through national surveys, which are very expensive and have significant lag time between data collection and published results.
In addition, these surveys suffer from well-known limitations that may underestimate substance use disorders (e.g., these surveys usually do not include people experiencing homelessness or people who are incarcerated, both populations disproportionately affected by substance use disorders). In these national surveys “treatment need” is defined as the percentage of all individuals who had a substance use disorder in the past year; “demand for treatment” is defined as the percentage of those individuals who received treatment or who reported seeking treatment in the past year. Historically, there has been a wide gap between “treatment need” and “demand for treatment,” though this gap differs by type of substance, and the demand for treatment is oftentimes not measured in regions, states, or counties. In this study, researchers examined characteristics and trends in help-seeking through analyzing Google searches for addiction treatment from 2010 to 2020 at the national and state level. Results could inform an approach to assess treatment demand at scale that is geographically precise, low-cost, and near real-time.
HOW WAS THIS STUDY CONDUCTED?
Researchers examined geographic characteristics and trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine through Google queries from 2010 to 2020 at the national and state level. They validated these help-seeking internet searches by looking at their associations with unmet treatment need, treatment seeking, and health harms related to unmet treatment demand.
The researchers obtained Google queries from 2010 to 2020 originating from the United States. Relevant queries for help-seeking included the terms “quit,” “stop,” “rehab(s),” “rehabilitation,” “treatment(s),” “help,” or “detox.” To separate the Google queries into help-seeking for specific substances, researchers used the terms “alcohol,” “alcoholic,” and “alcoholism” for alcohol, “cannabis,” and “marijuana” for cannabis, “opioid(s),” “heroin,” “fentanyl,” “oxycontin,” “oxycodone,” “codeine,” “hydrocodone,” and “morphine” for opioids, “methamphetamine” and “meth” for methamphetamine, and “cocaine” for cocaine.
For example, an internet search for “Where can I get help for alcoholism?” would count as an alcohol help-seeking search. An internet search for “Where can I find rehabilitation for meth use” would count as a methamphetamine help-seeking search. An internet search for “I need treatment for cocaine” would count as a cocaine help-seeking search. Notably, general searches without specifying a specific substance (e.g., “where can I find addiction treatment”) were not included in these analyses. Given this term (“addiction treatment”) may be quite a popular search term and way that people search for treatment, this is likely a limitation. The search query data were obtained for each calendar year from 2010 to 2020 and this data was measured at both the national and state level. Google queries were measured in “query fractions” – i.e., the number of addiction treatment searches per 1 million overall searches – which allowed for standardization over the time period and across states so that the results were comparable.
The researchers validated the help-seeking internet search results using several national databases that also contain state-level data. The ultimate aim of the study was to see if there were associations between the help-seeking internet searches and unmet treatment need, treatment seeking, and substance use-related health harms. Examining these associations would provide insight on the treatment capacity and the demand for treatment. For example, if there is sufficient treatment capacity, a strong, positive association between help-seeking internet searches and treatment admissions would be expected. On the contrary, if there is excess demand for treatment (i.e., more demand for treatment that can be provided by the treatment system), a weaker association between help-seeking internet searches and treatment admissions but a stronger association with unmet treatment need and drug-related health harms would be expected.
The researchers used the National Survey on Drug Use and Health to examine the associations between google search queries for help-seeking and unmet treatment need, defined as needing but not receiving treatment for both alcohol and illicit drug use, accounting statistically for differences between states, correlations between successive time points, and correlations between data points collected in the same states over different years. They used the Treatment Episodes Dataset: Admissions to examine the associations between help-seeking and treatment admissions for alcohol, cannabis, opioids, methamphetamine, and cocaine. They used the Healthcare Cost and Utilization Project and CDC WONDER to examine the associations between help-seeking and opioid-related harm, including nonfatal opioid overdoses and opioid overdose deaths.
WHAT DID THIS STUDY FIND?
Help-seeking internet searches for most substances were relatively stable over time though alcohol increased.
Generally, help-seeking internet searches were highest at the beginning of the study period in 2010 for all substances except alcohol, decreased the following year, then remained relatively stable before slightly increasing in 2020. Help-seeking internet searches for alcohol decreased in 2011 and then rose steadily from 2012 to 2020. In 2020, the year that the COVID-19 pandemic began, help-seeking internet searches for alcohol increased 21% compared to the previous year and help-seeking internet searches for methamphetamine increased 23%.
There were important state differences in help-seeking internet searches.
There was large variability in help-seeking estimated from Google searches for methamphetamine by state and generally low variability for alcohol, meaning that there were large differences between states for methamphetamine help-seeking but much smaller differences among states for alcohol help-seeking. Between 2010 and 2020, West Virginia (opioids and methamphetamine), Delaware (opioids), New Mexico (methamphetamine), and Connecticut (cocaine) repeatedly showed high levels of treatment demand for specific substances, as measured by help-seeking internet searches.
Help-seeking internet searches marked real-world behaviors for some but not all substances.
Findings showed a positive association between help-seeking internet searches and needing but not receiving treatment for alcohol use, a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, and a positive association between help-seeking internet searches and opioid overdose deaths. For example, a 1 unit increase in the query fraction for help-seeking internet searches for methamphetamine and opioid use were associated with a 26% and 12% increase, respectively, in the expected rate of treatment episodes. On the contrary, a 1 unit increase in the query fraction for help-seeking internet searches for opioid use was associated with a 11% increase in the expected overdose mortality counts.
WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?
This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by reliable federally funded databases.
It is unclear why help-seeking internet searches were generally highest in 2010 for most substances, subsequently decreased in 2011, then remained relatively stable throughout the study period up to 2020. There were two clear patterns: alcohol help-seeking internet searches rose steadily from 2011 to 2020 and both alcohol and methamphetamine help-seeking internet searches increased markedly in 2020 during the COVID-19 pandemic. Interestingly, as the opioid crisis worsened from 2010 to 2020, help-seeking internet searches for opioids remained relatively stable.
The demand for treatment, as measured by national surveys like the National Survey on Drug Use and Health, assumes that all individuals with a substance use disorder “need treatment.” However, substance use disorder is a spectrum disease with different levels of severity (mild, moderate, and severe). Most individuals with a substance use disorder have a mild disorder and many experts concur that only individuals with moderate to severe substance use disorders will seek treatment. In addition, individuals with certain types of substances use disorders, like opioid use disorders, when they are in the moderate to severe range, may be more inclined to seek treatment due to the uncomfortable experience of opioid withdrawal.
There may be some value in using the millions of help-seeking internet searches for addiction treatment in a way that informs treatment demand and capacity surveillance nationally and by state, especially given its low cost and near real-time surveillance potential, although more research is needed.
Some findings were similar to known data and trends, suggesting potential for analyzing help-seeking internet searches. There were differences among states, with some states known to have extensive substance issues (e.g., opioids in West Virginia and methamphetamine in New Mexico) showing up as outliers in these substance-specific help-seeking internet searches as would be expected. In addition, there was large state-to-state variability in methamphetamine help-seeking internet searches and the lowest state-to-state variability in alcohol help-seeking internet searches, reflecting the fact that methamphetamine is endemic to certain regions of the United States where it is more available and accessible whereas alcohol being universally accessible is more widespread. Lastly, increases in methamphetamine and alcohol help-seeking internet searches during 2020, the year the COVID-19 pandemic began, may serve as a signal for increased substance use and progressing negative impact and severity. At the very least, analyzing help-seeking internet searches could complement other information sources for informing allocation of resources and public health campaigns.
When the researchers validated their findings with federally-funded databases with national and state-level data, the findings were mixed and suggest that only help-seeking internet searches for some substances may be markers of treatment demand and treatment capacity.
If there is sufficient treatment capacity, one would expect to see a strong, positive association between help-seeking internet searches and treatment admissions. If there is excess demand for treatment, one would expect to see a strong, positive association between help-seeking internet searches and both unmet treatment need and drug-related harm and a weak association with treatment admissions. These relationships were not clear in this study. For example, even though alcohol help-seeking internet searches rose during the study period, there was a very small statistically significant association with unmet treatment need and no significant association with treatment admissions. This would seem to suggest that there may not be sufficient capacity for alcohol treatment. However, many individuals with alcohol problems may go to Alcoholics Anonymous or other mutual health organizations, which would not be captured in the treatment admission data. In addition, these are mere associations and the effect sizes are extremely small, so more research is needed to determine the value of using internet search behavior to estimate addiction treatment demand.
Finally, there was a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, suggesting that these searches could be used to assess treatment demand for these substances. However, there was also a positive association between help-seeking internet searches for opioids and opioid overdose deaths, which contrast with the researchers’ hypothesis that increased treatment admissions will decrease substance-related harm. More research is needed to tease out how help-seeking internet searches are related to real-world outcomes and how these associations could inform treatment demand at the national and state levels.
Some people may search for treatment on the internet using search terms that do not specific a substance, such as “addiction treatment” or “addiction rehab,” which would not be captured by the researcher’s protocol.
Some states had missing data for some substance-specific help-seeking internet searches because Google will only report search query data beyond a minimal threshold.
There is likely selection bias in the internet search data because not all individuals have equal access to the internet.
Help-seeking internet searches may have been out of curiosity or may have represented a family member or loved one seeking help for another individual. In addition, one individual may exhaustively search the internet for treatment, while another may search for treatment once, but each help-seeking internet search query is a proxy for one individual.
Aggregate Google search data is limited to designated marketing areas, which do not align with the jurisdiction levels of public health departments. This limits help-seeking internet searches as a surveillance tool.
There are many limitations in the databases used to validate the help-seeking internet searches. For example, the Treatment Episode Data Set only includes admissions to facilities that receive public funding, and the National Survey for Drug Use does not include institutionalized populations, such as people in jails and prisons.
Some important search terms for certain substances may have been left out and searches in other languages were not included.
BOTTOM LINE
In this study, researchers examined state-level and national trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine from 2010 to 2020 at the national and state level. This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by federally funded databases. While results are intriguing, and more research is certainly needed regarding the ultimate utility of this kind of approach in aiding public health, these Google search analyses may at the very least help augment national public health strategies to address alcohol and other drug use disorders in the United States.
For individuals and families seeking recovery: The internet is an increasingly common source used to find addiction treatment in the United States. Reliable sources include the Addiction Treatment Locator, Assessment, and Standards Platform (ATLAS), created by Shatterproof, and FindTreatment.Gov, a website run by the federal government.
For treatment professionals and treatment systems: Although single facilities may have a good idea of their treatment capacity and the demand for treatment in their area, better tools are needed to inform capacity and demand in local communities and regions as well as the nation as a whole. Help-seeking internet searches have the potential to enhance surveillance systems at the national, state, or county level regarding treatment needs in the population although more research is needed. These searches are low cost and provide near real-time data and could be one of many information sources that informs allocation of treatment resources and public health campaigns.
For scientists:Given the low cost and near real-time data that could be realized with an accurate surveillance tool to measure treatment demand in the United States, more research may be warranted to study the potential of help-seeking internet searches for addiction treatment. This may include using monthly data instead of yearly data, using county-level data instead of state-level data, adding additional covariates of interest, and replicating this study in another country. A body of literature beyond this one study could inform a surveillance system that prospectively assesses treatment demand.
For policy makers: Analyzing help-seeking internet searches for addiction treatment has the potential to aid stakeholders in creating surveillance tools to gauge treatment demand as well as informing allocation of resources and public health campaigns, although more research is needed. These searches are particularly attractive because of their low cost and near real-time data. At the very least, these searches may be beneficial in complementing other information sources for informing policies and interventions. More funding to determine the value of help-seeking internet searches as a surveillance tool could be fruitful.
Demand for addiction treatment, also known as help-seeking, is currently measured in the United States through national surveys, which are very expensive and have significant lag time between data collection and published results.
In addition, these surveys suffer from well-known limitations that may underestimate substance use disorders (e.g., these surveys usually do not include people experiencing homelessness or people who are incarcerated, both populations disproportionately affected by substance use disorders). In these national surveys “treatment need” is defined as the percentage of all individuals who had a substance use disorder in the past year; “demand for treatment” is defined as the percentage of those individuals who received treatment or who reported seeking treatment in the past year. Historically, there has been a wide gap between “treatment need” and “demand for treatment,” though this gap differs by type of substance, and the demand for treatment is oftentimes not measured in regions, states, or counties. In this study, researchers examined characteristics and trends in help-seeking through analyzing Google searches for addiction treatment from 2010 to 2020 at the national and state level. Results could inform an approach to assess treatment demand at scale that is geographically precise, low-cost, and near real-time.
HOW WAS THIS STUDY CONDUCTED?
Researchers examined geographic characteristics and trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine through Google queries from 2010 to 2020 at the national and state level. They validated these help-seeking internet searches by looking at their associations with unmet treatment need, treatment seeking, and health harms related to unmet treatment demand.
The researchers obtained Google queries from 2010 to 2020 originating from the United States. Relevant queries for help-seeking included the terms “quit,” “stop,” “rehab(s),” “rehabilitation,” “treatment(s),” “help,” or “detox.” To separate the Google queries into help-seeking for specific substances, researchers used the terms “alcohol,” “alcoholic,” and “alcoholism” for alcohol, “cannabis,” and “marijuana” for cannabis, “opioid(s),” “heroin,” “fentanyl,” “oxycontin,” “oxycodone,” “codeine,” “hydrocodone,” and “morphine” for opioids, “methamphetamine” and “meth” for methamphetamine, and “cocaine” for cocaine.
For example, an internet search for “Where can I get help for alcoholism?” would count as an alcohol help-seeking search. An internet search for “Where can I find rehabilitation for meth use” would count as a methamphetamine help-seeking search. An internet search for “I need treatment for cocaine” would count as a cocaine help-seeking search. Notably, general searches without specifying a specific substance (e.g., “where can I find addiction treatment”) were not included in these analyses. Given this term (“addiction treatment”) may be quite a popular search term and way that people search for treatment, this is likely a limitation. The search query data were obtained for each calendar year from 2010 to 2020 and this data was measured at both the national and state level. Google queries were measured in “query fractions” – i.e., the number of addiction treatment searches per 1 million overall searches – which allowed for standardization over the time period and across states so that the results were comparable.
The researchers validated the help-seeking internet search results using several national databases that also contain state-level data. The ultimate aim of the study was to see if there were associations between the help-seeking internet searches and unmet treatment need, treatment seeking, and substance use-related health harms. Examining these associations would provide insight on the treatment capacity and the demand for treatment. For example, if there is sufficient treatment capacity, a strong, positive association between help-seeking internet searches and treatment admissions would be expected. On the contrary, if there is excess demand for treatment (i.e., more demand for treatment that can be provided by the treatment system), a weaker association between help-seeking internet searches and treatment admissions but a stronger association with unmet treatment need and drug-related health harms would be expected.
The researchers used the National Survey on Drug Use and Health to examine the associations between google search queries for help-seeking and unmet treatment need, defined as needing but not receiving treatment for both alcohol and illicit drug use, accounting statistically for differences between states, correlations between successive time points, and correlations between data points collected in the same states over different years. They used the Treatment Episodes Dataset: Admissions to examine the associations between help-seeking and treatment admissions for alcohol, cannabis, opioids, methamphetamine, and cocaine. They used the Healthcare Cost and Utilization Project and CDC WONDER to examine the associations between help-seeking and opioid-related harm, including nonfatal opioid overdoses and opioid overdose deaths.
WHAT DID THIS STUDY FIND?
Help-seeking internet searches for most substances were relatively stable over time though alcohol increased.
Generally, help-seeking internet searches were highest at the beginning of the study period in 2010 for all substances except alcohol, decreased the following year, then remained relatively stable before slightly increasing in 2020. Help-seeking internet searches for alcohol decreased in 2011 and then rose steadily from 2012 to 2020. In 2020, the year that the COVID-19 pandemic began, help-seeking internet searches for alcohol increased 21% compared to the previous year and help-seeking internet searches for methamphetamine increased 23%.
There were important state differences in help-seeking internet searches.
There was large variability in help-seeking estimated from Google searches for methamphetamine by state and generally low variability for alcohol, meaning that there were large differences between states for methamphetamine help-seeking but much smaller differences among states for alcohol help-seeking. Between 2010 and 2020, West Virginia (opioids and methamphetamine), Delaware (opioids), New Mexico (methamphetamine), and Connecticut (cocaine) repeatedly showed high levels of treatment demand for specific substances, as measured by help-seeking internet searches.
Help-seeking internet searches marked real-world behaviors for some but not all substances.
Findings showed a positive association between help-seeking internet searches and needing but not receiving treatment for alcohol use, a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, and a positive association between help-seeking internet searches and opioid overdose deaths. For example, a 1 unit increase in the query fraction for help-seeking internet searches for methamphetamine and opioid use were associated with a 26% and 12% increase, respectively, in the expected rate of treatment episodes. On the contrary, a 1 unit increase in the query fraction for help-seeking internet searches for opioid use was associated with a 11% increase in the expected overdose mortality counts.
WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?
This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by reliable federally funded databases.
It is unclear why help-seeking internet searches were generally highest in 2010 for most substances, subsequently decreased in 2011, then remained relatively stable throughout the study period up to 2020. There were two clear patterns: alcohol help-seeking internet searches rose steadily from 2011 to 2020 and both alcohol and methamphetamine help-seeking internet searches increased markedly in 2020 during the COVID-19 pandemic. Interestingly, as the opioid crisis worsened from 2010 to 2020, help-seeking internet searches for opioids remained relatively stable.
The demand for treatment, as measured by national surveys like the National Survey on Drug Use and Health, assumes that all individuals with a substance use disorder “need treatment.” However, substance use disorder is a spectrum disease with different levels of severity (mild, moderate, and severe). Most individuals with a substance use disorder have a mild disorder and many experts concur that only individuals with moderate to severe substance use disorders will seek treatment. In addition, individuals with certain types of substances use disorders, like opioid use disorders, when they are in the moderate to severe range, may be more inclined to seek treatment due to the uncomfortable experience of opioid withdrawal.
There may be some value in using the millions of help-seeking internet searches for addiction treatment in a way that informs treatment demand and capacity surveillance nationally and by state, especially given its low cost and near real-time surveillance potential, although more research is needed.
Some findings were similar to known data and trends, suggesting potential for analyzing help-seeking internet searches. There were differences among states, with some states known to have extensive substance issues (e.g., opioids in West Virginia and methamphetamine in New Mexico) showing up as outliers in these substance-specific help-seeking internet searches as would be expected. In addition, there was large state-to-state variability in methamphetamine help-seeking internet searches and the lowest state-to-state variability in alcohol help-seeking internet searches, reflecting the fact that methamphetamine is endemic to certain regions of the United States where it is more available and accessible whereas alcohol being universally accessible is more widespread. Lastly, increases in methamphetamine and alcohol help-seeking internet searches during 2020, the year the COVID-19 pandemic began, may serve as a signal for increased substance use and progressing negative impact and severity. At the very least, analyzing help-seeking internet searches could complement other information sources for informing allocation of resources and public health campaigns.
When the researchers validated their findings with federally-funded databases with national and state-level data, the findings were mixed and suggest that only help-seeking internet searches for some substances may be markers of treatment demand and treatment capacity.
If there is sufficient treatment capacity, one would expect to see a strong, positive association between help-seeking internet searches and treatment admissions. If there is excess demand for treatment, one would expect to see a strong, positive association between help-seeking internet searches and both unmet treatment need and drug-related harm and a weak association with treatment admissions. These relationships were not clear in this study. For example, even though alcohol help-seeking internet searches rose during the study period, there was a very small statistically significant association with unmet treatment need and no significant association with treatment admissions. This would seem to suggest that there may not be sufficient capacity for alcohol treatment. However, many individuals with alcohol problems may go to Alcoholics Anonymous or other mutual health organizations, which would not be captured in the treatment admission data. In addition, these are mere associations and the effect sizes are extremely small, so more research is needed to determine the value of using internet search behavior to estimate addiction treatment demand.
Finally, there was a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, suggesting that these searches could be used to assess treatment demand for these substances. However, there was also a positive association between help-seeking internet searches for opioids and opioid overdose deaths, which contrast with the researchers’ hypothesis that increased treatment admissions will decrease substance-related harm. More research is needed to tease out how help-seeking internet searches are related to real-world outcomes and how these associations could inform treatment demand at the national and state levels.
Some people may search for treatment on the internet using search terms that do not specific a substance, such as “addiction treatment” or “addiction rehab,” which would not be captured by the researcher’s protocol.
Some states had missing data for some substance-specific help-seeking internet searches because Google will only report search query data beyond a minimal threshold.
There is likely selection bias in the internet search data because not all individuals have equal access to the internet.
Help-seeking internet searches may have been out of curiosity or may have represented a family member or loved one seeking help for another individual. In addition, one individual may exhaustively search the internet for treatment, while another may search for treatment once, but each help-seeking internet search query is a proxy for one individual.
Aggregate Google search data is limited to designated marketing areas, which do not align with the jurisdiction levels of public health departments. This limits help-seeking internet searches as a surveillance tool.
There are many limitations in the databases used to validate the help-seeking internet searches. For example, the Treatment Episode Data Set only includes admissions to facilities that receive public funding, and the National Survey for Drug Use does not include institutionalized populations, such as people in jails and prisons.
Some important search terms for certain substances may have been left out and searches in other languages were not included.
BOTTOM LINE
In this study, researchers examined state-level and national trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine from 2010 to 2020 at the national and state level. This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by federally funded databases. While results are intriguing, and more research is certainly needed regarding the ultimate utility of this kind of approach in aiding public health, these Google search analyses may at the very least help augment national public health strategies to address alcohol and other drug use disorders in the United States.
For individuals and families seeking recovery: The internet is an increasingly common source used to find addiction treatment in the United States. Reliable sources include the Addiction Treatment Locator, Assessment, and Standards Platform (ATLAS), created by Shatterproof, and FindTreatment.Gov, a website run by the federal government.
For treatment professionals and treatment systems: Although single facilities may have a good idea of their treatment capacity and the demand for treatment in their area, better tools are needed to inform capacity and demand in local communities and regions as well as the nation as a whole. Help-seeking internet searches have the potential to enhance surveillance systems at the national, state, or county level regarding treatment needs in the population although more research is needed. These searches are low cost and provide near real-time data and could be one of many information sources that informs allocation of treatment resources and public health campaigns.
For scientists:Given the low cost and near real-time data that could be realized with an accurate surveillance tool to measure treatment demand in the United States, more research may be warranted to study the potential of help-seeking internet searches for addiction treatment. This may include using monthly data instead of yearly data, using county-level data instead of state-level data, adding additional covariates of interest, and replicating this study in another country. A body of literature beyond this one study could inform a surveillance system that prospectively assesses treatment demand.
For policy makers: Analyzing help-seeking internet searches for addiction treatment has the potential to aid stakeholders in creating surveillance tools to gauge treatment demand as well as informing allocation of resources and public health campaigns, although more research is needed. These searches are particularly attractive because of their low cost and near real-time data. At the very least, these searches may be beneficial in complementing other information sources for informing policies and interventions. More funding to determine the value of help-seeking internet searches as a surveillance tool could be fruitful.
Demand for addiction treatment, also known as help-seeking, is currently measured in the United States through national surveys, which are very expensive and have significant lag time between data collection and published results.
In addition, these surveys suffer from well-known limitations that may underestimate substance use disorders (e.g., these surveys usually do not include people experiencing homelessness or people who are incarcerated, both populations disproportionately affected by substance use disorders). In these national surveys “treatment need” is defined as the percentage of all individuals who had a substance use disorder in the past year; “demand for treatment” is defined as the percentage of those individuals who received treatment or who reported seeking treatment in the past year. Historically, there has been a wide gap between “treatment need” and “demand for treatment,” though this gap differs by type of substance, and the demand for treatment is oftentimes not measured in regions, states, or counties. In this study, researchers examined characteristics and trends in help-seeking through analyzing Google searches for addiction treatment from 2010 to 2020 at the national and state level. Results could inform an approach to assess treatment demand at scale that is geographically precise, low-cost, and near real-time.
HOW WAS THIS STUDY CONDUCTED?
Researchers examined geographic characteristics and trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine through Google queries from 2010 to 2020 at the national and state level. They validated these help-seeking internet searches by looking at their associations with unmet treatment need, treatment seeking, and health harms related to unmet treatment demand.
The researchers obtained Google queries from 2010 to 2020 originating from the United States. Relevant queries for help-seeking included the terms “quit,” “stop,” “rehab(s),” “rehabilitation,” “treatment(s),” “help,” or “detox.” To separate the Google queries into help-seeking for specific substances, researchers used the terms “alcohol,” “alcoholic,” and “alcoholism” for alcohol, “cannabis,” and “marijuana” for cannabis, “opioid(s),” “heroin,” “fentanyl,” “oxycontin,” “oxycodone,” “codeine,” “hydrocodone,” and “morphine” for opioids, “methamphetamine” and “meth” for methamphetamine, and “cocaine” for cocaine.
For example, an internet search for “Where can I get help for alcoholism?” would count as an alcohol help-seeking search. An internet search for “Where can I find rehabilitation for meth use” would count as a methamphetamine help-seeking search. An internet search for “I need treatment for cocaine” would count as a cocaine help-seeking search. Notably, general searches without specifying a specific substance (e.g., “where can I find addiction treatment”) were not included in these analyses. Given this term (“addiction treatment”) may be quite a popular search term and way that people search for treatment, this is likely a limitation. The search query data were obtained for each calendar year from 2010 to 2020 and this data was measured at both the national and state level. Google queries were measured in “query fractions” – i.e., the number of addiction treatment searches per 1 million overall searches – which allowed for standardization over the time period and across states so that the results were comparable.
The researchers validated the help-seeking internet search results using several national databases that also contain state-level data. The ultimate aim of the study was to see if there were associations between the help-seeking internet searches and unmet treatment need, treatment seeking, and substance use-related health harms. Examining these associations would provide insight on the treatment capacity and the demand for treatment. For example, if there is sufficient treatment capacity, a strong, positive association between help-seeking internet searches and treatment admissions would be expected. On the contrary, if there is excess demand for treatment (i.e., more demand for treatment that can be provided by the treatment system), a weaker association between help-seeking internet searches and treatment admissions but a stronger association with unmet treatment need and drug-related health harms would be expected.
The researchers used the National Survey on Drug Use and Health to examine the associations between google search queries for help-seeking and unmet treatment need, defined as needing but not receiving treatment for both alcohol and illicit drug use, accounting statistically for differences between states, correlations between successive time points, and correlations between data points collected in the same states over different years. They used the Treatment Episodes Dataset: Admissions to examine the associations between help-seeking and treatment admissions for alcohol, cannabis, opioids, methamphetamine, and cocaine. They used the Healthcare Cost and Utilization Project and CDC WONDER to examine the associations between help-seeking and opioid-related harm, including nonfatal opioid overdoses and opioid overdose deaths.
WHAT DID THIS STUDY FIND?
Help-seeking internet searches for most substances were relatively stable over time though alcohol increased.
Generally, help-seeking internet searches were highest at the beginning of the study period in 2010 for all substances except alcohol, decreased the following year, then remained relatively stable before slightly increasing in 2020. Help-seeking internet searches for alcohol decreased in 2011 and then rose steadily from 2012 to 2020. In 2020, the year that the COVID-19 pandemic began, help-seeking internet searches for alcohol increased 21% compared to the previous year and help-seeking internet searches for methamphetamine increased 23%.
There were important state differences in help-seeking internet searches.
There was large variability in help-seeking estimated from Google searches for methamphetamine by state and generally low variability for alcohol, meaning that there were large differences between states for methamphetamine help-seeking but much smaller differences among states for alcohol help-seeking. Between 2010 and 2020, West Virginia (opioids and methamphetamine), Delaware (opioids), New Mexico (methamphetamine), and Connecticut (cocaine) repeatedly showed high levels of treatment demand for specific substances, as measured by help-seeking internet searches.
Help-seeking internet searches marked real-world behaviors for some but not all substances.
Findings showed a positive association between help-seeking internet searches and needing but not receiving treatment for alcohol use, a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, and a positive association between help-seeking internet searches and opioid overdose deaths. For example, a 1 unit increase in the query fraction for help-seeking internet searches for methamphetamine and opioid use were associated with a 26% and 12% increase, respectively, in the expected rate of treatment episodes. On the contrary, a 1 unit increase in the query fraction for help-seeking internet searches for opioid use was associated with a 11% increase in the expected overdose mortality counts.
WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?
This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by reliable federally funded databases.
It is unclear why help-seeking internet searches were generally highest in 2010 for most substances, subsequently decreased in 2011, then remained relatively stable throughout the study period up to 2020. There were two clear patterns: alcohol help-seeking internet searches rose steadily from 2011 to 2020 and both alcohol and methamphetamine help-seeking internet searches increased markedly in 2020 during the COVID-19 pandemic. Interestingly, as the opioid crisis worsened from 2010 to 2020, help-seeking internet searches for opioids remained relatively stable.
The demand for treatment, as measured by national surveys like the National Survey on Drug Use and Health, assumes that all individuals with a substance use disorder “need treatment.” However, substance use disorder is a spectrum disease with different levels of severity (mild, moderate, and severe). Most individuals with a substance use disorder have a mild disorder and many experts concur that only individuals with moderate to severe substance use disorders will seek treatment. In addition, individuals with certain types of substances use disorders, like opioid use disorders, when they are in the moderate to severe range, may be more inclined to seek treatment due to the uncomfortable experience of opioid withdrawal.
There may be some value in using the millions of help-seeking internet searches for addiction treatment in a way that informs treatment demand and capacity surveillance nationally and by state, especially given its low cost and near real-time surveillance potential, although more research is needed.
Some findings were similar to known data and trends, suggesting potential for analyzing help-seeking internet searches. There were differences among states, with some states known to have extensive substance issues (e.g., opioids in West Virginia and methamphetamine in New Mexico) showing up as outliers in these substance-specific help-seeking internet searches as would be expected. In addition, there was large state-to-state variability in methamphetamine help-seeking internet searches and the lowest state-to-state variability in alcohol help-seeking internet searches, reflecting the fact that methamphetamine is endemic to certain regions of the United States where it is more available and accessible whereas alcohol being universally accessible is more widespread. Lastly, increases in methamphetamine and alcohol help-seeking internet searches during 2020, the year the COVID-19 pandemic began, may serve as a signal for increased substance use and progressing negative impact and severity. At the very least, analyzing help-seeking internet searches could complement other information sources for informing allocation of resources and public health campaigns.
When the researchers validated their findings with federally-funded databases with national and state-level data, the findings were mixed and suggest that only help-seeking internet searches for some substances may be markers of treatment demand and treatment capacity.
If there is sufficient treatment capacity, one would expect to see a strong, positive association between help-seeking internet searches and treatment admissions. If there is excess demand for treatment, one would expect to see a strong, positive association between help-seeking internet searches and both unmet treatment need and drug-related harm and a weak association with treatment admissions. These relationships were not clear in this study. For example, even though alcohol help-seeking internet searches rose during the study period, there was a very small statistically significant association with unmet treatment need and no significant association with treatment admissions. This would seem to suggest that there may not be sufficient capacity for alcohol treatment. However, many individuals with alcohol problems may go to Alcoholics Anonymous or other mutual health organizations, which would not be captured in the treatment admission data. In addition, these are mere associations and the effect sizes are extremely small, so more research is needed to determine the value of using internet search behavior to estimate addiction treatment demand.
Finally, there was a positive association between help-seeking internet searches and treatment admissions for methamphetamine and opioid use, suggesting that these searches could be used to assess treatment demand for these substances. However, there was also a positive association between help-seeking internet searches for opioids and opioid overdose deaths, which contrast with the researchers’ hypothesis that increased treatment admissions will decrease substance-related harm. More research is needed to tease out how help-seeking internet searches are related to real-world outcomes and how these associations could inform treatment demand at the national and state levels.
Some people may search for treatment on the internet using search terms that do not specific a substance, such as “addiction treatment” or “addiction rehab,” which would not be captured by the researcher’s protocol.
Some states had missing data for some substance-specific help-seeking internet searches because Google will only report search query data beyond a minimal threshold.
There is likely selection bias in the internet search data because not all individuals have equal access to the internet.
Help-seeking internet searches may have been out of curiosity or may have represented a family member or loved one seeking help for another individual. In addition, one individual may exhaustively search the internet for treatment, while another may search for treatment once, but each help-seeking internet search query is a proxy for one individual.
Aggregate Google search data is limited to designated marketing areas, which do not align with the jurisdiction levels of public health departments. This limits help-seeking internet searches as a surveillance tool.
There are many limitations in the databases used to validate the help-seeking internet searches. For example, the Treatment Episode Data Set only includes admissions to facilities that receive public funding, and the National Survey for Drug Use does not include institutionalized populations, such as people in jails and prisons.
Some important search terms for certain substances may have been left out and searches in other languages were not included.
BOTTOM LINE
In this study, researchers examined state-level and national trends in help-seeking internet searches for alcohol, cannabis, opioids, methamphetamine, and cocaine from 2010 to 2020 at the national and state level. This study found differences across states and over time in help-seeking internet searches for addiction treatment in the United States and found important substance-specific differences in the association between these searches and unmet treatment need, treatment seeking, and health harms related to unmet treatment demand as measured by federally funded databases. While results are intriguing, and more research is certainly needed regarding the ultimate utility of this kind of approach in aiding public health, these Google search analyses may at the very least help augment national public health strategies to address alcohol and other drug use disorders in the United States.
For individuals and families seeking recovery: The internet is an increasingly common source used to find addiction treatment in the United States. Reliable sources include the Addiction Treatment Locator, Assessment, and Standards Platform (ATLAS), created by Shatterproof, and FindTreatment.Gov, a website run by the federal government.
For treatment professionals and treatment systems: Although single facilities may have a good idea of their treatment capacity and the demand for treatment in their area, better tools are needed to inform capacity and demand in local communities and regions as well as the nation as a whole. Help-seeking internet searches have the potential to enhance surveillance systems at the national, state, or county level regarding treatment needs in the population although more research is needed. These searches are low cost and provide near real-time data and could be one of many information sources that informs allocation of treatment resources and public health campaigns.
For scientists:Given the low cost and near real-time data that could be realized with an accurate surveillance tool to measure treatment demand in the United States, more research may be warranted to study the potential of help-seeking internet searches for addiction treatment. This may include using monthly data instead of yearly data, using county-level data instead of state-level data, adding additional covariates of interest, and replicating this study in another country. A body of literature beyond this one study could inform a surveillance system that prospectively assesses treatment demand.
For policy makers: Analyzing help-seeking internet searches for addiction treatment has the potential to aid stakeholders in creating surveillance tools to gauge treatment demand as well as informing allocation of resources and public health campaigns, although more research is needed. These searches are particularly attractive because of their low cost and near real-time data. At the very least, these searches may be beneficial in complementing other information sources for informing policies and interventions. More funding to determine the value of help-seeking internet searches as a surveillance tool could be fruitful.
