Untangling Treatment Effect From Patient Factors: The Challenge of the Use of Observational Studies to Investigate the Impact of Positive Airway Pressure Therapy on Medical Costs

Abstract

FOR RELATED ARTICLE, SEE PAGE 1543OSA meaningfully impairs quality of life, and positive airway pressure (PAP) treatment has been rigorously demonstrated to improve symptoms such as sleepiness.1Pack A.I. Magalang U.J. Singh B. Kuna S.T. Keenan B.T. Maislin G. Randomized clinical trials of cardiovascular disease in obstructive sleep apnea: understanding and overcoming bias.Sleep. 2021; 44: 1-11Crossref Scopus (40) Google Scholar Further, OSA perturbs sleep physiology, causing long-term harm including cardiovascular disease and death.1Pack A.I. Magalang U.J. Singh B. Kuna S.T. Keenan B.T. Maislin G. Randomized clinical trials of cardiovascular disease in obstructive sleep apnea: understanding and overcoming bias.Sleep. 2021; 44: 1-11Crossref Scopus (40) Google Scholar Given its positive impact on OSA, it is plausible that PAP therapy might reduce medical costs related to OSA sequelae. Over the last 2 decades, there have been considerable efforts to evaluate whether PAP therapy reduces medical costs.2Chhatre S. Chang Y.H.A. Gooneratne N.S. Kuna S. Strollo P. Jayadevappa R. Association between adherence to continuous positive airway pressure treatment and cost among Medicare enrollees.Sleep. 2020; 43: 1-10Crossref Scopus (9) Google Scholar, 3Derose S.F. Zhou H. Huang B.Z. Manthena P. Hwang D. Shi J.M. Does providing positive airway pressure for sleep apnea change health care utilization?.Med Care. 2018; 56: 901-907Crossref PubMed Scopus (7) Google Scholar, 4Kirsch D.B. Yang H. Maslow A.L. Stolzenbach M. McCall A. Association of positive airway pressure use with acute care utilization and costs.J Clin Sleep Med. 2019; 15: 1243-1250Crossref PubMed Scopus (8) Google Scholar, 5Wickwire E.M. Bailey M.D. Somers V.K. et al.CPAP adherence is associated with reduced inpatient utilization among older adult Medicare beneficiaries with pre-existing cardiovascular disease.J Clin Sleep Med. 2022; 18: 39-45Crossref PubMed Scopus (0) Google Scholar Conducting randomized controlled trials to answer this question is not feasible, given ethical concerns and difficulty in achieving optimal long-term PAP adherence. Therefore, observational study designs are commonly employed. FOR RELATED ARTICLE, SEE PAGE 1543 In this issue of CHEST, An et al6An J. Glick H.A. Sawyer A.M. et al.Association between positive airway pressure adherence and health care costs among individuals with obstructive sleep apnea.Chest. 2023; 163: 1543-1554Google Scholar used data on 543 patients with OSA who were prescribed PAP to assess health care costs associated with PAP adherence. Patients were followed for 3 years; the 1-year trend in costs prior to PAP was also evaluated. Patients were stratified into high (25%), moderate (22%), and low (52%) groups using PAP machine-measured adherence. The high adherence group included patients with an average of ≥ 4 hours PAP use per night for every 6-month interval. A pronounced increase in costs in the first 6 months after PAP initiation was most prominent in the low adherence group. Subsequently, costs decreased in all groups. The average cost was significantly lower in the high vs low adherence group during the first (P = .002) and second (P = .02) 6-month periods only. In the primary analysis, the high adherence group had adjusted 6-month total and non-sleep-related costs that were lower by $832 and $1,107 respectively, compared to the low adherence group. The resource category that contributed greatest to this cost difference was hospitalization, which was $1905 lower in the high vs low adherence group. These findings concur with those of prior studies that indicate that, although costs are not reduced in all patients who are prescribed PAP, those most adherent to PAP incur lower costs.2Chhatre S. Chang Y.H.A. Gooneratne N.S. Kuna S. Strollo P. Jayadevappa R. Association between adherence to continuous positive airway pressure treatment and cost among Medicare enrollees.Sleep. 2020; 43: 1-10Crossref Scopus (9) Google Scholar, 3Derose S.F. Zhou H. Huang B.Z. Manthena P. Hwang D. Shi J.M. Does providing positive airway pressure for sleep apnea change health care utilization?.Med Care. 2018; 56: 901-907Crossref PubMed Scopus (7) Google Scholar, 4Kirsch D.B. Yang H. Maslow A.L. Stolzenbach M. McCall A. Association of positive airway pressure use with acute care utilization and costs.J Clin Sleep Med. 2019; 15: 1243-1250Crossref PubMed Scopus (8) Google Scholar, 5Wickwire E.M. Bailey M.D. Somers V.K. et al.CPAP adherence is associated with reduced inpatient utilization among older adult Medicare beneficiaries with pre-existing cardiovascular disease.J Clin Sleep Med. 2022; 18: 39-45Crossref PubMed Scopus (0) Google Scholar The current and prior observational studies that have assessed the economic impact of PAP adherence are susceptible to biases that threaten the validity of their findings.7Shrank W.H. Patrick A.R. Brookhart M.A. Healthy user and related biases in observational studies of preventive interventions: a primer for physicians.J Gen Intern Med. 2011; 26: 546-550Crossref PubMed Scopus (355) Google Scholar,8Kapur V.K. Psaty B.M. Obtaining valid estimates of the effect of CPAP therapy: reducing healthy adherer and other biases in observational studies.Chest. 2022; 161: 1444-1445Abstract Full Text Full Text PDF PubMed Scopus (0) Google Scholar Biases such as the healthy adherer effect and residual confounding cause estimates of effect to deviate erroneously from the null. Healthy adherer bias occurs when PAP-adherent patients engage more often in healthy behaviors, such as adherence to medication and following recommended guidance on screening, vaccination, and lifestyle behaviors that lead to improved health. Benefits ascribed to high PAP adherence may derive from these other behaviors. The magnitude of this bias can be large: a meta-analysis that compared placebo control subjects in randomized controlled trials with high vs low adherence found high placebo adherence was associated with a 44% reduction in mortality rates.9Simpson S.H. Eurich D.T. Majumdar S.R. et al.A meta-analysis of the association between adherence to drug therapy and mortality.BMJ. 2006; 333: 15Crossref PubMed Google Scholar To their credit, An et al6An J. Glick H.A. Sawyer A.M. et al.Association between positive airway pressure adherence and health care costs among individuals with obstructive sleep apnea.Chest. 2023; 163: 1543-1554Google Scholar attempted to minimize healthy adherer bias by adjusting for adherence to medications (antihypertensive, antidiabetic, and statin medications) and missed clinic visits in their regression models. As expected, adherence to these medications and clinic show rates were highest in the high adherer group. Interestingly, individuals in the low adherence group were less likely to take these medications. This highlights a potential weakness in relying on adherence to medications as a proxy for healthy behavior: only those who receive and fill a prescription will have a value for the measure. In addition, those who do not fill a received prescription could be conflated erroneously with those who never received a prescription. A similar issue may impact the use of clinic show rates: exclusion of people who should visit the clinic but never make an appointment. To better address healthy adherer bias, future studies could incorporate additional measures of healthy behavior such as receipt of immunizations, appropriate screening (eg, colonoscopy, mammogram), and receipt of annual wellness visits. Accounting for other behaviors that are not captured adequately in the medical record, such as diet, physical activity, and regular alcohol use, would also be useful. Residual confounding comes from patient characteristics associated with PAP adherence and health care costs that are not measured and therefore cannot be adjusted for in a regression model. The authors had access to multiple potential confounding clinical variables including BMI, apnea-hypopnea index, and medical comorbidity, which is a strength of the study. OSA symptom burden (eg, sleepiness) is another clinically important characteristic that is worthy of inclusion. To reduce confounding because of prior levels of health care use, the authors adjusted for log-transformed 6-month costs prior to PAP initiation. Given seasonality in health care use, controlling for costs during the 12 months before PAP initiation, ideally on a more granular level (eg, every month or every 3 months) in future studies may be helpful. Reliance on routine clinical care to identify OSA and treat patients can introduce bias if PAP adherence status is associated with a differential likelihood of receiving medical care after OSA diagnosis. Higher costs immediately after PAP initiation that were observed most prominently in the low adherence group suggest that patients who did not engage previously with the health care system may have begun receiving care for neglected needs. Although this bias most obviously impacts studies of costs, increased interaction with the health care system also increases opportunities for receiving diagnoses when disease outcomes are of interest. An in-depth exploration of health care use immediately after PAP initiation could identify whether short-term changes in health care seeking behavior were occurring. Exclusion of this period from analysis in future studies could mitigate the impact of this bias. The exploration by An et al6An J. Glick H.A. Sawyer A.M. et al.Association between positive airway pressure adherence and health care costs among individuals with obstructive sleep apnea.Chest. 2023; 163: 1543-1554Google Scholar of the impact of PAP therapy on medical costs produced findings that concur with prior observational studies. Their work augments preexisting literature by incorporating direct measures of PAP adherence and addressing some limitations that are intrinsic to this study design. Notably, the investigators commendably attempted to reduce bias because of confounding and healthy adherer effect. Confounding related to differential engagement with the health care system is an overlooked source of bias that remains to be evaluated and addressed. To satisfy skeptics, future work will need to incorporate improved strategies that address threats to validity that are posed by observational study designs. The authors have reported to CHEST the following: V. K. K. is a board member of the American Academy of Sleep Medicine. J. S. A.’s institution has received research funding from the American Academy of Sleep Medicine Foundation, Department of Defense, Merck, ResMed, and the ResMed Foundation. Disclaimer: The views expressed are his own and do not necessarily represent the views of the American Academy of Sleep Medicine. Other contributions: V. K. K. contributed to this article in his personal capacity. Association Between Positive Airway Pressure Adherence and Health Care Costs Among Individuals With OSACHESTVol. 163Issue 6PreviewIn participants with OSA, better PAP adherence was associated with significantly lower health care costs over 3 years. Findings support the importance of strategies to enhance long-term PAP adherence. Full-Text PDF