Use of Electronic Health Information Technology in a National Sample of Hospitals That Provide Specialty Substance Use Care

Substance use disorders cause significant morbidity and mortality (1). Individuals receiving specialty substance use disorder care in hospital settings are often among those with the most severe conditions and associated health complications. Furthermore, the risk of drug interactions between alcohol and drugs and prescribed medications (2–4)—including increased risk of death (5)—is considerable. These safety risks are heightened when medications used to treat substance use disorders are not listed in the medical record (6). Indeed, managing these risks requires not only accurate documentation in the health record but also robust health care coordination and communication among providers caring for patients in the hospital and those providing postdischarge care.

Electronic health records (EHRs) offer opportunities to improve the quality and safety of health care. They have been demonstrated to significantly reduce medication errors and adverse drug events, improve guideline adherence, and facilitate communication with patients (7–11). Recent evidence has also shown that they are associated with reduced hospital mortality (12). Additionally, use of electronic health information exchange (HIE) has been associated with improved health care quality and safety and reductions in unnecessary utilization and costs (13); use also facilitates timely information exchange to providers at the next level of care.

As of 2015, 84% of acute care hospitals had adopted at least basic EHR functionality in at least one unit of a hospital (14). As of 2017, 88% of acute care hospitals electronically sent HIE to providers at the next level of care (15). Recognizing the importance of HIE for the quality and safety of health care, the 21st Century Cures Act requires the Office of the National Coordinator (ONC) for Health Information Technology to promulgate regulations to further support the exchange of health information. In May 2020, the ONC published its final rule to advance this goal (16).

However, use of basic EHR functionality and electronic HIE by hospital-based specialty substance use programs may be lower than the national estimates of EHR adoption and use of HIE in acute care hospitals. One reason may be that providers of specialty substance use disorder treatment are subject to a federal health privacy law (42 Code of Federal Regulations Part 2) that is considerably stricter than the Health Insurance Portability and Accountability Act (HIPAA), and EHRs have limited ability to manage these additional privacy requirements. Furthermore, Part 2–covered programs may be disproportionately located in psychiatric hospitals—hospitals identified as having much lower basic EHR adoption rates (14).

Information about EHR adoption at acute care hospitals—specifically, at least basic EHR functionality adoption and use of electronic HIE—is derived from hospital responses to the American Hospital Association’s Annual Survey Information Technology (IT) Supplement. The functional features that constitute basic EHR adoption were determined by expert consensus (17), are considered the standard definition of basic adoption (18–21), and have been used by the ONC in its national estimates (14, 22). The IT Supplement also provides the ONC with national estimates of hospital-based electronic HIE. However, it cannot provide estimates for specific programs within hospitals, because it asks hospital respondents to report on adoption “in at least one unit” of the hospital. Thus, for programs within hospitals that are subject to additional federal privacy requirements (such as substance use disorder specialty programs), EHR adoption or use of HIE in these programs may not match the EHR adoption or use of HIE at the hospital overall.

The aim of this study was to estimate the use of basic EHR technology and electronic HIE by using data from the Substance Abuse and Mental Health Administration’s (SAMHSA) National Survey of Substance Abuse Treatment Services (N-SSATS) and determine hospital characteristics that predict use of EHR technology and electronic HIE within these settings. We hypothesized that the rate of basic EHR functionality adoption and electronic HIE use in these programs would be lower than national estimates for hospitals overall.

Methods

Data Source

The N-SSATS is an annual census of all specialty providers of substance use treatment services in the United States. It is administered to specialty providers at all levels of care (e.g., inpatient, residential, partial hospital, and outpatient) and, in the case of hospitals, among acute care general hospitals, psychiatric hospitals, and other specialty hospitals (e.g., alcoholism and maternity). The 2017 survey is the most recent N-SSATS that asked respondents about EHR adoption and electronic HIE use; it had an overall response rate of 89.2%. The study reported here was not human subjects research, and thus it did not require approval by an institutional review board.

Primary Outcomes of Interest

We examined two primary outcomes reported by hospital-based substance use disorder programs in the 2017 N-SSATS: adoption of basic EHR functional features and use of electronic HIE. The N-SSATS EHR adoption questions corresponded with those of the ONC but did not entirely match. Therefore, we used the N-SSATS responses to create an indicator of basic EHR adoption that approximated the ONC’s definition as closely as possible. We defined basic EHR features in the N-SSATS as electronic-computer sources to accomplish the following five clinical tasks: assessment, client progress monitoring, discharge, issue and receive lab results, and medication prescribing and dispensing (see Table 1 for a comparison of the functional features examined based on the ONC definition and the N-SSATS items). In contrast, the definition of HIE in the N-SSATS is the same as the ONC’s (i.e., “Do you use HIE to send client health and/or treatment information to providers or sources outside of your organization?”).

A further difference between the ONC and the N-SSATS is that the ONC definition asks about adoption in “at least one unit of the hospital,” whereas the N-SSATS asks respondents to “indicate if staff members routinely use computer or electronic resources, paper only, or a combination of both to accomplish their work” in their program. In the N-SSATS, electronic resources are defined as “includ[ing] tools such as electronic health records and web portals,” and paper documents are defined as “PDFs, scanned documents, or e-fax.” To approximate the ONC’s definition of “in at least one unit of the hospital,” we defined adoption of basic EHR or electronic HIE as occurring if a hospital used “any” computer-electronic source (i.e., computer-electronic only or both paper and electronic-computer) in a substance use disorder program. We used “any” for our definition, because a hospital may have more than one substance use disorder treatment program or unit (e.g., in different levels of care) and it is possible that EHR features were adopted in one but not (or not yet) in all of the programs (hence a mix of paper and computer use).

Explanatory Variables

Explanatory variables examined included hospital type (general acute care hospitals, psychiatric hospitals, and other specialty hospitals), hospital profit status (nonprofit, for profit, and government hospitals), whether the program accepts Medicaid, and whether it provides medication treatment for opioid or alcohol use disorders (methadone, naltrexone, buprenorphine, acamprosate, or disulfiram). Hospital characteristics, such as hospital type and profit status, have been shown to be associated with EHR adoption and HIE use in previous studies (including in behavioral health hospital settings) (18, 20, 23, 24). We examined whether a program accepts Medicaid, because Medicaid is a predominant payer of substance use disorder services (25). Finally, we examined whether a program provides medications to treat alcohol or opioid use disorders, because of the important role of EHRs in health care safety and quality associated with medication use and prescribing and because access to medication treatment is a measure of the quality of a substance use disorder program in general. We also created a variable that described the highest level of substance use disorder care provided by the organization (inpatient, residential, intensive outpatient or partial hospitalization, and regular outpatient), because health information technology (HIT) is likely to be most beneficial in managing care for patients with complex medical needs and who are the sickest.

Statistical Analysis

We calculated descriptive statistics of the characteristics of hospital-based substance use disorder programs and conducted bivariate analyses of associations between the explanatory variables and each primary outcome. We then fit multivariable logistic regression models for our two outcomes. In the model of electronic HIE, we also included the indicator for EHR adoption as a predictor. To address respondents with missing outcome information, we followed a missing value approach used in prior research on EHR adoption (24). Specifically, we estimated the probability of each respondent’s having a missing outcome and then weighted respondents without missing data by 1/(1−probability of having missing outcome). Using this approach, respondents with nonmissing outcomes compensated for similar respondents with missing outcomes and thus reduced the risk of bias. We estimated predicted probabilities (PPs) for significant predictors following each model. We tested the sensitivity of our specifications of the outcomes by fitting models where missing values on the outcome variables were forced into the “no” category.

Results

There were 1,024 hospitals with substance use disorder programs in the 2017 N-SSATS. In unadjusted analyses, we excluded 130 (13%) hospitals from the EHR analyses and 185 hospitals (18%) from the HIE analyses because of missing outcome information. Of the 894 facilities with outcome information that were included in the EHR adoption analysis, two-thirds (N=606, 68%) met our study criteria for having adopted basic EHR functionality (Table 2). Of the 839 hospitals with outcome information on use of HIE, 598 (71%) reported using electronic HIE. (A CONSORT diagram in an online supplement to this article shows differences in characteristics by missingness.) Compared with hospital-based substance use disorder programs that did not adopt basic EHR functionality, those that did were more likely to be in acute care general hospitals and nonprofit hospitals and to provide substance use disorder medications and offer routine outpatient care as their highest level of care. The patterns for use of electronic HIE were similar across facility types and ownership categories, as well as in having routine outpatient treatment or intensive outpatient or partial hospitalization as the highest level of care. When we limited the analysis to hospital-based programs in acute care general hospitals, EHR and electronic HIE adoption were 75% and 74%, respectively (data not shown).

Programs in psychiatric hospitals had lower odds of basic EHR adoption, compared with acute care general hospitals (odds ratio [OR]=0.49 [Table 3]; psychiatric hospitals, PP=0.56; acute care general hospitals, PP=0.73 [Table 4]). Compared with nonprofit hospitals, programs in for-profit or government-owned hospitals also had lower odds of basic EHR adoption (for profit, OR=0.23; PP=42; government owned, OR=0.52; PP=0.56). Greater odds of basic EHR adoption were associated with provision of substance use disorder medications (OR=1.95; provides such medications, PP=0.70; does not provide such medications, PP=0.57). Lower odds of basic EHR adoption was associated with having intensive outpatient or partial hospitalization as the highest level of care, compared with having routine outpatient care as the highest level (OR=0.49; intensive outpatient or partial hospitalization, PP=0.62; routine outpatient care, PP=0.75). In the logistic model predicting use of HIE, only adoption of basic EHR functionality was associated with electronic HIE use (OR=4.73; has basic EHR, PP=0.81; does not have basic EHR, PP=0.48) (Table 4). The results of the sensitivity analysis were qualitatively unchanged from our primary analysis.

Discussion

In this national sample of hospital-based substance use disorder treatment programs, we found that adoption of basic EHR functionality and electronic HIE to send information to outside providers lagged behind EHR adoption and HIE use in acute care hospitals (14, 15). About two-thirds (68%) of the hospital-based substance use treatment programs reported adopting basic EHR functionality, compared with 84% of acute care hospitals, and 71% of the substance use treatment programs reported sending electronic HIE to outside providers, compared with 88% of acute care hospitals. Prior work has found that psychiatric hospitals have low rates of basic EHR adoption (14, 24). Because the programs in our sample were disproportionately located in psychiatric hospitals, it is not surprising that an overall estimate of basic EHR adoption among hospital-based substance use disorder programs was lower than the ONC estimates. However, even when we limited our analyses to hospital-based programs in acute care general hospitals, EHR adoption and HIE use were lower (75% and 74%, respectively), compared with previously reported rates of basic EHR adoption and HIE use in acute care hospitals overall (84% and 88%, respectively) (14, 15).

Our findings suggest that compared with adoption at the overall hospital level, there are unique barriers to EHR adoption and HIE use among hospital-based substance use disorder programs. Patients with substance use disorders are known to experience stigma related to their illness, including by health care professionals (26). Possibly these experiences have led some programs to be reluctant to implement EHRs. Another potential barrier, as noted above, may be that the federal 42 CFR Part 2 privacy requirements are more stringent than HIPAA. Additionally, some states have privacy laws that are more stringent than Part 2 (27). Part 2 has been criticized as making it difficult for substance use disorder treatment information to be documented in EHRs (28, 29). In February 2017, and more recently (July 2020), SAMHSA updated the Part 2 regulations to address these concerns (30, 31). Our data, from the 2017 N-SSATS, is unlikely to reflect any changes that may have occurred because of these updates to Part 2. Future research is needed to understand whether, or to what extent, the updates to Part 2 have enabled broader adoption of EHRs and HIE use in substance use disorder treatment settings.

Compared with substance use disorder programs based in nonprofit hospitals, those based in for-profit and government hospitals were less likely to have adopted basic EHR functionality, which reflects similar patterns in EHR adoption more broadly (18). This finding is notable considering the rise of for-profit substance use disorder treatment programs in the United States. Although most concerns about for-profit substance use disorder treatment programs have focused on lack of evidence-based treatments, elements such as use of a basic EHR and electronic HIE are foundational to a wide range of quality assessment and improvement activities. However, we did not find ownership to be a significant predictor of HIE use, mirroring a similar finding from a recent study of electronic HIE use among inpatient psychiatric units of acute care hospitals (23). Differences by ownership among hospital-based providers, therefore, appear to be related to initial investment in HIT infrastructure and not necessarily to the use of existing HIT for information sharing at discharge.

Hospital-based substance use disorder treatment programs that provided medications for alcohol or opioid use disorders were more likely than those that did not provide such medications to have adopted basic EHR functionality. Given the medication safety benefits of EHRs, this is a welcome finding. Still, a note of caution in interpreting these results is that the N-SSATS asks only about a single component of EHR-based medication information functionality—that is, prescribing or dispensing. An important EHR safety and quality feature that is also a part of the ONC definition of basic EHR functionality, but is not part of the information collected in the N-SSATS, is the presence of a medication list. Surprisingly, we did not find that organizations offering higher intensity levels of substance use disorder care were more likely than those offering routine outpatient care to adopt basic EHR or electronic HIE. In fact, providers offering intensive outpatient treatment or partial hospitalization were less likely than those offering only routine outpatient care to adopt basic her. Further research is needed to understand reasons for this finding.

The strongest predictor of electronic HIE use in hospital-based substance use disorder programs was having adopted basic EHR functionality. This finding is not surprising, given that EHRs would be a primary vehicle by which patient electronic health information could be exchanged. Nevertheless, it highlights that if policy makers want to increase opportunities to improve the quality and safety of care enabled by electronic HIE in substance use disorder programs, more efforts will be needed to encourage EHR adoption among specialty substance use disorder treatment providers. Additionally, EHR adoption by psychiatric hospitals faces challenges. Psychiatric hospitals were ineligible for the incentives available to general acute care hospitals from the federal Health Information Technology for Economic and Clinical Health (HITECH) Act. The HITECH Act, which provided hospitals with financial assistance and incentives to adopt EHRs, was credited as a success in spurring HIT adoption by eligible hospitals (19). No such similar program has been specifically enacted for psychiatric hospitals, but the 2018 Support Act could potentially provide the means to investigate such a policy. It allows the Center for Medicare and Medicaid Innovation to test incentive demonstration models for EHR adoption in behavioral health settings (32). As yet, no demonstration program of this type has been announced.

An important limitation of this study is that the N-SSATS HIT items did not entirely align with the definitions of basic EHR adoption and HIE use specified by the ONC. To the extent that this study misclassified whether a program had adopted basic EHR technology or used electronic HIE, we expect our estimates are an overestimate, compared with the ONC’s. Another limitation is that we were limited to hospital characteristics available in the N-SSATS. The N-SSATS database does not include hospital identifiers. Therefore, we were unable to supplement the data with other publicly available hospital and facility data, including information previously associated with EHR adoption and HIE use in hospitals and psychiatric units—e.g., hospital size (20, 21, 23) or participation in an accountable care organization (23). Finally, as we noted above, these data were from 2017—the most current year available—which might not reflect current rates of EHR and HIE adoption.

The differences between the N-SSATS and the ONC definitions are problematic from a health policy perspective. The N-SSATS is a missed opportunity to better compare adoption of electronic technology in hospital-based substance use disorder treatment programs with adoption in acute care hospitals overall. Recent work by members of our team has identified a similar issue in the Centers for Medicare and Medicaid Services’ (CMS’s) Inpatient Psychiatric Facility Quality Reporting Program (23). In addition to N-SSATS, SAMHSA’s National Mental Health Services Survey has similar limitation for measuring EHR adoption and HIE use. Given the importance of HIT in the quality and safety of health care, important future policy efforts should not only encourage adoption in behavioral health treatment settings but also apply consistent definitions of adoption to behavioral health settings, as are used elsewhere in health care settings and by policy makers.

Conclusions

The adoption of HIT that could improve the quality and safety of hospital-based substance use disorder specialty care lags behind adoption in hospital-based treatment settings more broadly. This finding was not entirely accounted for by HIT adoption in psychiatric hospitals (which is known to lag behind adoption in acute care hospitals), but it appears that there are additional barriers to EHR and HIE use in substance use disorder treatment programs based in acute care general hospital settings (possibly related to 42 CFR Part 2). Understanding the extent to which HIT adoption lags in specialty behavioral health programs would be improved by better alignment across the ONC, CMS, and SAMHSA surveys that examine such adoption.