Prescription Opioid Abuse: Gleaning insights from hospital and vital records data

Reka Sundaram-Stukel, Ousmane Diallo, Benjamin Wiseman, Richard E. Miller

Abstract


ObjectiveIn this paper we used hospital charges to assess costs incurred dueto prescription drug/opioid hospitalizationsIntroductionThere is a resurgence in the need to evaluate the economic burdenof prescription drug hospitalizations in the United States. We used theWisconsin 2014 Hospital Discharge data to examine opioid relatedhospitalization incidence and costs. Fentanyl, a powerful syntheticopioid, is frequently being used for as an intraoperative agent inanesthesia, and post-operative recovery in hospitals. According to a2013 study, synthetic Fentanyl is 40 times more potent than heroinand other prescription opioids; the strength of Fentanyl leads tosubstantial hospitalizations risks. Since, 1990 it has been availablewith a prescription in various forms such as transdermal patches orlollipops for treatment of serious chronic pain, most often prescribedfor late stage cancer patients. There have been reported fatal overdosesassociated with misuse of prescription fentanyl. In Wisconsin numberof total opioid related deaths increased by 51% from 2010 to 2014with the number of deaths involving prescription opioids specificallyincreased by 23% and number of deaths involving heroin increasedby 192%. We hypothesized that opioids prescription drugs, as a proxyof Fentanyl use, result in excessive health care costs.MethodsOpioid hospitalizations was defined as any mention of the ICD9codes (304,305) in any diagnostic field or the mention of (:E935.09) onthe first listed E-code. Our analysis used the Heckman 2-stage model,a method often used by Economists in absence of randomized controltrials. In presence of unobserved choice, for example opioid relatedhospitalizations, there usually is a correlation between error in anunderlying function (fentanyl prescription) and an estimated function(hospital charges) that introduces a selection bias. Heckman treats thiscorrelation between errors as an omitted variable bias. Therefore, weestimate a Heckman two step model using hospitalization: where theselection function is the probability of being hospitalized for syntheticopioid via logistic regression. Finally, we estimate the hospitalcharges realized if the patient was given opioids.ResultsMale patients are significantly more likely to be hospitalized foropioids than are female patients; while white patients are significantlymore likely to be admitted for opioid usage than other racialgroups. We also find that comorbid factors, such as mental health,significantly impact hospital charges associated with opioid use. Wefind that persons with private health insurance are associated withhigher rates of opioid use.ConclusionsUsing a Heckman two step approach we show that comorbidconditions such as mental health, Hepatitis C, injuries, etc significantlyaffect hospital charges associated with hospitalization. We usethese findings to explore the impact of the 2013 rule mandatingdoctors share opioid prescription information on the incidence ofopioid related death and hospital charges associated with opioidprescriptions. This work is policy relevant because alternatives toopioid prescription such as meditation, pain management therapiesmay be relevant.

Full Text:

PDF


DOI: https://doi.org/10.5210/ojphi.v9i1.7723



Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org