Outpatient Provider Contact Prior to Unintentional Opioid Overdose Among VHA Service Users

Fatal unintentional overdose, also referred to as death by “poisoning,” has increased substantially over the past decade, becoming the most common injury-related cause of death among adults in the United States (1). In recent years, pharmaceutical opioids have become the substance most often implicated in these overdose deaths, and prescription opioid–related deaths are now more common than deaths related to cocaine, heroin, and psychostimulants combined (2).

A number of studies have examined individual clinical and demographic risk factors for unintentional opioid overdose. Individuals with psychiatric and substance use disorder diagnoses, particularly opioid use disorders, have higher rates of unintentional nonfatal drug overdose (3). Misuse of prescription opioid medications is also common among those who die from overdose (4,5). The risk of fatal overdose has been shown to be associated with the total daily dose of prescription opioid (6,7), and individuals prescribed high-dose opioids have more comorbid pain and other general medical conditions, as well as substance abuse and other psychiatric conditions (8).

Although there is increasing recognition of unintentional prescription opioid overdoses as a rapidly growing national problem, few interventions are known to reduce risk of overdose among patient populations, aside from efforts that seek to improve prescribing practices. Recent research and implementation efforts have been based on potential strategies for intervening with individuals who have been identified as at risk for prescription opioid overdose or for improving the likelihood of survival if an overdose occurs (9–12).

Questions about how, when, and where such interventions can be targeted within health systems remain unanswered. Treatment data from individuals identified as users of a specific health system who died of a prescription opioid overdose can help improve our understanding of prevention opportunities. Thus, to inform the design of opioid overdose prevention interventions, we examined the types of treatment settings visited by patients prior to opioid overdose death and the temporal proximity of these visits to death. We also examined how demographic and clinical characteristics, such as psychiatric and pain conditions, differed among patients on the basis of treatment setting. We used data from the Veterans Health Administration (VHA), which serves a national population at higher risk of overdose (13). Use of data from an integrated health system allowed comparison of different outpatient treatment settings through a national electronic medical records system (13).

Methods

Study data were obtained from the Department of Veterans Affairs (VA) National Patient Care Database (NPCD) and the National Death Index (NDI). Study methods were approved by the Ann Arbor VA’s Institutional Review Board.

Sample

To identify VA patients who overdosed from fiscal year (FY) 2004 to FY 2007, we first identified all individuals who used VA services during this period on the basis of treatment records in the NPCD. We then examined whether these individuals had any record of contact with a VHA treatment provider in FY 2008 or FY 2009 and, thus, were known to be alive through the end of the observation period (end of FY 2007). NDI searches conducted for the remaining individuals with no VA service utilization in FY 2008 or FY 2009 identified 1,813 unintentional opioid overdose deaths from FY 2004 to FY 2007.

Cause of Death

The NDI includes national data regarding dates and causes of death for all U.S. residents, derived from death certificates filed in state vital statistics offices. Fatal unintentional poisoning was defined using ICD-10 codes X42, X44, Y12, and Y14 (14). We included deaths ruled unintentional or indeterminate in intent, consistent with prior studies in the study population (6).

The measure of death due to prescription opioid overdose was also based on the T codes included in NDI records. We included codes representing unintentional overdose on any prescription opioid (including T codes 40.2, 40.3, and 40.4). These criteria encompassed overdoses due to nonsynthetic and semisynthetic opioids (for example, codeine, morphine, oxycodone, hydrocodone, oxymorphone, and hydromorphone; code 40.2) and other opioids (that is, methadone and other synthetic opioids). Heroin (T code 40.1; found for 3.9% of this sample) and other substances may also have been involved, but a prescription opioid was involved in all of the overdoses included.

To serve as comparison groups, a random sample of all veterans who died of any cause and another random sample of veterans who died from injury death (excluding unintentional overdose) between FY 2004 and FY 2007 and who had used VA services in the two years before death were also analyzed.

Demographic Information

Demographic information available for each patient included age in years (18–44, 45–64, and ≥65), race (white, black, and unknown or other), and ethnicity (Hispanic ethnicity or other). Reliable data on other demographic characteristics (for example, employment and salary) were not available for this sample.

Diagnoses

Substance use disorders, other psychiatric conditions, pain, and other medical diagnoses were all based on ICD-9-CM diagnostic codes (14), reflecting clinical diagnoses made by VA treatment providers during clinical encounters in the year prior to unintentional overdose death. Specific substance use disorders examined were any diagnoses of intoxication, withdrawal, abuse, or dependence involving alcohol, cocaine, cannabis, opioids, benzodiazepines, and multiple substances or other. The multiple substances or other category included individuals with an ICD-9 clinical diagnosis of “polysubstance abuse” or “polysubstance dependence,” as well as individuals with a rarer substance use disorder diagnosis (for example, inhalant abuse). Participants could have diagnoses of multiple substance use disorders. Presence or absence of the following psychiatric diagnoses during one year before overdose death was also examined: major depression, schizophrenia, bipolar disorder I or II, posttraumatic stress disorder, and other anxiety disorders. In addition, we included multiple medical conditions, including pain disorders. These specific conditions were selected because of the frequency for which opioids are prescribed for their treatment. Other common medical conditions (that is, arthritis, cardiovascular disease, and chronic obstructive pulmonary disease) were also included.

Treatment Utilization

Clinic stop codes in the NPCD were examined to generate indicators of care within the 12 months before death. The following indicators were utilized to reflect outpatient care received within seven, 30, and 90 days before death and within the year before death: any substance use disorder treatment, any mental health treatment, any mental health or substance use disorder treatment, pain clinic treatment, primary care treatment, and any other medical treatment. The majority of the visits in the “other medical treatment” category were outpatient medical care visits in the following settings: specialty outpatient clinics, admission or screening, and telephone triage. Specialty outpatient settings included any nonprimary care clinical settings, such as cardiology and orthopedics clinics, where a patient would have had direct contact with a provider for clinical reasons. In addition, in analyses comparing demographic, clinical, and treatment characteristics across settings, the specific setting of care of the final visit was examined and categorized into the following mutually exclusive categories: pain, specialty outpatient, primary care, other medical, mental health, and substance use disorder clinics. Some patients received more than one type of care on the date of the last visit. In order to have mutually exclusive categories, these patients were coded as having their last visit in whichever setting was least common in the sample overall. In addition, we developed measures for number of days between final treatment contact and date of overdose death, whether a patient filled an opioid prescription on the day of the final visit, and whether a patient was in opioid substitution treatment.

Analyses

We examined the treatment received in the year prior to overdose death, subdivided by type of treatment and by the time proximity of final treatment contact, among individuals with any VHA contact within two years prior to death. Further analyses focused on patients whose final contact prior to overdose death was in an outpatient setting. In this subsample, we first compared demographic characteristics, psychiatric conditions, and general medical conditions across outpatient treatment locations by using chi square tests. We also looked at differences in the percentages of patients who filled an opioid prescription on the day of the final visit and three months, six months, and two years prior to death and the proportions of patients in opioid substitution treatment across outpatient contact settings prior to overdose by using chi square tests. Finally, using analysis of variance, we examined the average number of days across contact settings between the final visit and overdose death.

Results

This study included 1,813 patients who died of a prescription opioid overdose in FY 2004 to FY 2007 and who had medical contact within two years of their death. Of this sample, 1,457 (80%) patients were last seen in a VHA outpatient setting within one year prior to death. This subgroup was considered to have engaged in treatment recently and was the subsample included in detailed analysis of demographic and clinical factors.

Among patients who were last seen in outpatient settings (N=1,457), more were seen in the month before death in mental health clinics (26%) and primary care clinics (31%) than in substance use disorder clinics (8%) or pain clinics (3%) (Table 1). In comparison, in the random sample of 111,999 patients who died from any cause in the same period, 44% (N=48,901) were last seen in primary care, compared with only 5% (N=5,281) last seen in mental health clinics and .4% (N=448) last seen in substance use disorder clinics. In the random sample of 5,075 patients who died from other injury-related causes (excluding unintentional overdose), 43% (N=2,186) were last seen in primary care, 11% (N=549) were last seen in mental health clinics, and 2% (N=85) were last seen in substance disorder clinics.

Table 2 reports the demographic characteristics for the 1,457 patients who were last seen in outpatient settings prior to overdose death, with patients categorized by final treatment site. Consistent with the general VHA patient population, 92% of the sample was male. Gender did not significantly differ among clinic types. Distribution of patients in different age groups (p=.003) and race (p=.049) differed by clinic type; patients last seen in a pain clinic tended to be younger and were more likely to be white, compared with patients last seen in all other settings.

Table 3 displays the frequency of general medical and psychiatric conditions by treatment setting. Overall, a substantial proportion of patients last seen in medical, mental health, and substance use disorder clinics had pain and other medical conditions, although not surprisingly, pain conditions were even more prevalent among patients last seen in a pain clinic. A smaller proportion of patients with any psychiatric condition were last seen in medical settings than in mental health or in substance use disorder clinics. Patients with substance use disorders were more likely to have been last seen in a substance use disorder clinic, but the proportion of patients with substance use disorders was similar across other settings.

Only 24% of patients last seen in primary care, 5% of patients last seen in mental health clinics, and 2% of patients last seen in a substance use disorder clinic filled an opioid prescription on the day of their last outpatient visit prior to unintentional opioid overdose (Table 4). A total of 725 of the 1,457 patients filled an opioid prescription in the six months before death. The most common opioids filled in the year prior to death included oxycodone (N=372, 34%), hydrocodone (N=333, 31%), and morphine (N=250, 23%). In addition, 48% of the sample (N=702) filled a benzodiazepine prescription in the year prior to death. A significantly larger proportion of patients in substance use disorder clinics (33%) were in opioid substitution treatment compared with all other clinics.

Discussion

This study is the first to our knowledge to examine types of clinical contact prior to death from an unintentional prescription opioid overdose. Data from this national cohort of all patients seen in the VHA (defined as those who had received any care in the prior two years) who died from an overdose of prescription opioids showed that many were seen in outpatient clinic settings within 30 days of their overdose. This suggests that people who die from an overdose of prescription opioids are likely to have recently engaged in treatment; consequently, there is the potential in the outpatient context for targeting interventions to prevent prescription opioid overdoses.

It is noteworthy that the proportions of individuals last seen in a mental health clinic summed with those seen in a substance use disorder clinic were similar to the proportion seen in a primary care clinic. Furthermore, the proportions of patients with substance use disorders, pain, and other medical disorders were largely similar for those last seen in psychiatric treatment and in medical treatment settings. This similarity suggests that screening individuals on the basis of patient-level factors may be efficacious for identifying those at risk of prescription opioid overdose, regardless of specific treatment setting.

In addition, it appears that patients who died from unintentional opioid overdose were twice as likely to be last seen in mental health clinics and four times as likely to be last seen in substance use disorder clinics, compared with those who died from other injury-related causes. Although this finding is not surprising given the high percentages of substance use and other mental disorders in the unintentional-overdose sample, it emphasizes that outpatient psychiatric settings may provide an important opportunity for intervention to prevent death from unintentional overdose.

Despite accumulating data indicating that those who overdose have high rates of psychiatric and substance use disorder diagnoses, there are no known assessments to help identify those at risk of unintentional overdose and no prevention interventions focused on patients seen in mental health settings (15). Our data show that the large group of patients with psychiatric conditions was more likely to be last seen in psychiatric outpatient settings. Furthermore, most of these patients did not fill an opioid prescription in the VHA within the six months of their death, which suggests that provider-level prescribing interventions in the outpatient medical settings alone may not reach this group of patients with high prevalence of psychiatric disorders.

Even though in most cases of unintentional overdose in this study, patients did not obtain prescription opioids at mental health visits, there may still be an opportunity to intervene in this setting, especially because it has been shown that many of those who overdose use medications not prescribed to them (4,5). In addition, psychiatric providers are frequent prescribers of medications, such as benzodiazepines, that are commonly seen in overdose deaths and that may interact with opioids to increase the risk of overdose (16). In this sample of patients, benzodiazepines were commonly prescribed, with 48% of the sample filling a prescription in the year prior to death. Finally, psychiatric providers may be more specifically trained in psychosocial risk factors, particularly from their experiences in assessing suicide risk, which may help them better assess and discuss risk factors for unintentional overdose with patients.

These data also indicate that the majority of patients did not obtain prescription opioid medications from providers on their final visit prior to opioid overdose. In fact, among the 1,457 patients, 14% filled a prescription for an opioid on their last visit. Among those last seen in mental health clinics, only 5% filled a prescription for an opioid on their last visit. This finding suggests that it may be crucial to focus on screening for opioid overdose risk not only when a provider is prescribing or refilling an opioid medication but also as part of routine follow-up care for persons prescribed these agents. Thus screening may be appropriate at any outpatient visit for all patients prescribed opioids and could focus on factors that have been associated with increased overdose risk, such as dose of prescription opioids and presence of general medical and psychiatric conditions. This would be a step toward stratifying patients by risk scores into categories in order to target interventions more appropriately.

Finally, in contrast to psychiatric and other medical outpatient settings, specialty pain clinics seem to treat a small but unique subset of patients. Patients last seen in pain clinics had a different constellation of characteristics, including younger age, and as can be expected, a much higher likelihood of filling an opioid prescription on the day of the final visit. These patients also had a significantly shorter interval between their last visit and death. They were also more likely to have pain and to have lower rates of diagnosed substance use disorders. Patients in pain clinics may benefit from different screening and intervention approaches than patients in other types of clinics.

There were several limitations to this study. This was a study of patients actively receiving care in the VHA, which is one of the largest integrated health care systems in this country. Our results may not generalize to a different health care system, although the integrated nature of the VHA creates opportunities for developing and testing prevention and intervention strategies. These results may also not generalize to veterans who did not receive VHA care within a two-year time frame. In addition, there has been a significant shift in overdose mortality patterns in the past several decades (17). Our results did not examine temporal trends in patterns of treatment received prior to opioid overdose. Recent national data indicate that deaths from unintentional overdoses, particularly from prescription opioids, increased until 2010 (16,17), which may or may not have influenced the associations reported here.

Another limitation to this study is that we did not examine predictors of time to overdose death. In the future, if examined within specific treatment settings, such data could inform screening efforts.

Conclusions

Findings of this study indicate that outpatient clinics, particularly primary care and mental health clinics, may provide an opportunity to identify and intervene with patients at elevated risk of unintentional prescription opioid overdose. There is an increasing body of data on risk factors for overdose among patient populations, and the results of this study suggest that an important next step may be to create and implement risk stratification measures for outpatient clinical settings, such as primary care and mental health, to identify patients at risk of prescription opioid overdose. Furthermore, interventions developed to address unintentional overdose that are tailored to the primary care or mental health context could have a meaningful impact on unintentional overdoses among patients prescribed opioid medications.