Accuracy of patient-reported versus real-time electronic postoperative opioid use outcomes

OBJECTIVE:

Most studies examining home postoperative opioid consumption rely on retrospective patient self-reports^1,2; however, the accuracy of these reports is unknown. In addition, demonstrating the accuracy of opioid use reports is crucial to the interpretation of studies addressing the role of medical opioids in the current opioid crisis. In this study, we evaluated the accuracy of retrospective patient-reported analgesic use vs real-time analgesic use data obtained via electronic medication caps.

STUDY DESIGN:

The study was a secondary analysis of a randomized trial comparing home opioid use after cesarean delivery in women receiving discharge counseling regarding optimal analgesic use vs usual care.³ At discharge, participants received 30 hydrocodone (5 mg) tablets in bottles fitted with Bluetooth-enabled Smart Caps (Pillsy) that recorded dates and times of pill container access. Patients were instructed to take 1 pill every 4 hours as needed for pain. Smart Cap data were obtained via mobile applications installed on the participants’ smartphone that was synched to an electronic platform or directly from caps returned to study personnel. Each pill bottle access was recorded as a single consumed pill. Patients were educated to open the pill bottle twice if they took 2 pills at 1 time. To account for potential pill splitting, patients were asked during follow-up contact specifically how many times they took half a pill. The Smart Cap data were adjusted down 0.5 pill for the number of times the participant indicated they consumed half a pill. Patients were contacted by telephone or email on postoperative day 14 to report the number of opioid tablets used since discharge.

Statistical analyses included all participants with both Smart Cap and patient-reported data. An intraclass correlation (ICC) assessed the degree of correspondence between Smart Cap and patient-reported opioid use data. Patient-reported use was classified into 3 categories vis-à-vis Smart Cap data: (1) overreporting (using >2 tablets more than real-time data), (2) underreporting (using >2 tablets less), and (3) comparable reporting (using within 1 tablet). Likelihood chi-square tests assessed differences between these categories in demographic and clinical characteristics. The criterion for significance was a Bonferroni-corrected P value of <.05.

RESULTS:

Smart Cap and patient-reported data were available for 132 participants. Their mean age was 30.6±SD=5.3 years, and most participants were white (74%) and privately insured (66%) (Table). Of the 132 patients, 8 (6.1%) indicated taking more than 1 pill at a time, and 13 (9.6%) reported taking 0.5 pill between 2 and 5 times. Patient-reported use corresponded highly⁴ with Smart Cap data (ICC, 0.85; 95% confidence interval, 0.80–0.90; P<.001). Patient-reported use tended to align at 5-tablet intervals (ie, 5, 10, and 15 tablets), a distribution not observed for Smart Cap data, possibly reflecting the tendency to round numbers when recalling information (Figure). Comparable reporting of opioid use (50.8%) was the most common, followed by overreporting (32.6%) and underreporting of opioid use (16.7%). When reports were inaccurate, publicly insured or self-pay participants were more likely to underreport opioid use (55.0%), whereas privately insured participants were more likely to overreport opioid use (61.9%; P<.05). Participants overreporting opioid use had higher median inpatient opioid use than participants with comparable reports (25 vs 15 morphine milligram equivalents; P<.05). No other difference was significant, although participants with a diagnosis of depression trended (P=.052) toward more frequent overreporting of opioid use (Table).

TABLE.

Demographic and clinical characteristics by Smart Cap and self-report use category

Characteristic	Overall	Overreported	Comparable	Underreported	P value
Insurance					.042a
Private or military	80 (65.6)	26 (61.9)	45 (75.0)b	9 (45.0)c
Public or self-pay	42 (34.4)	16 (38.1)	15 (25.0)b	11 (55.0)c
Total number	122	42	60	20
Smoke during pregnancy					.642
No	120 (90.9)	38 (88.4)	61 (91.0)	21 (95.5)
Yes	12 (9.1)	5 (11.6)	6 (9.0)	1 (4.5)
Total number	132	43	67	22
Any comorbidities					.384
No	78 (60.0)	25 (59.5)	37 (56.1)	16 (72.7)
Yes	52 (40.0)	17 (40.5)	29 (43.9)	6 (27.3)
Total number	130	42	66	22
Any complications					.325
No	86 (65.6)	32 (74.4)	41 (61.2)	13 (61.9)
Yes	45 (34.4)	11 (25.6)	26 (38.8)	8 (38.1)
Total number	131	43	67	21
History depression					.052
No	107 (81.7)	30 (71.4)	56 (83.6)	21 (95.5)
Yes	24 (18.3)	12 (28.6)	11 (16.4)	1 (4.5)
Total number	131	42	67	22
Age (y)					.687
Median (IQR)	31.1 (27.0–35.0)	31.1 (27.0–34.0)	31.1 (27.0–35.0)	31.0 (22.0–34.0)
n	128	42	64	22
Overall pain after cesarean delivery					.236
Median (IQR)	5.0 (4.0–6.0)	5.0 (4.0–7.0)	4.0 (3.0–6.0)	5.0 (4.0–7.0)
n	131	43	66	22
Inpatient use (MME)					.009a
Median (IQR)	20.0 (7.0–30.0)	25.0 (15.0–35.0)b	15.0 (0.0–25.0)c	22.5 (10.0–45.0)
n	132	43	67	22

Values are number (percentage) unless indicated otherwise.

IQR, interquartile range; MME, morphine milligram equivalents.

^aStatistically significant values;

^b,cStatistically significantly different post-hoc Bonferroni-corrected P values of <.05.;

FIGURE.

Distributions of Pillsy and patient-reported opioid pill use after cesarean delivery (N=132)

CONCLUSION:

Overprescribing of postoperative opioids is well documented, and providers are increasingly seeking methods to align prescribing to patient need.⁵ Research in this arena hinges on the accuracy of opioid use assessment. Although self-reported opioid use is a simple, pragmatic, and inexpensive method to gather data, studies evaluating the accuracy of these data are lacking, ultimately leaving study findings relying on these methods open to questions. Real-time electronic monitoring of outpatient opioid use is more accurate but substantially more complex and expensive. This study fills a critical research gap, demonstrating high correspondence between patient recall of opioid use at 14 days after hospital discharge and contemporaneous electronic medication use monitoring. Self-reports, when inaccurate, tended more often to reflect overreporting of opioid use, similar to the broader medication adherence literature.⁶

The increased likelihood for publicly insured and self-pay participants to underreport opioid use requires further investigation. These findings were based on relatively small participant numbers and should be interpreted with caution. However, if replicated, these results could indicate that studies relying solely on self-reported opioid use may need to consider specific population characteristics as factors that potentially could alter interpretations of reported results. In previous research, publicly insured and self-pay patients have displayed postoperative opioid use patterns that diverge from those with private insurance, including higher opioid consumption after cesarean delivery² and other surgical procedures^7–9 and higher reports of receiving too few opioids after surgery.^2,10 Although this study fills a critical gap in the pain management literature, several limitations must be addressed. Here, patients were female, relatively young, and undergoing cesarean delivery, with no patient enrolled who had used opioids within 7 days preoperatively. It should also be noted that historical policies criminalizing substance use, including use of opioid analgesics, may influence a woman’s willingness to self-report opioid use.^11–13 The extent to which the current findings will generalize to other populations is not yet known. In addition, this cross-sectional study analyzed opioid use for only 2 weeks postoperatively. The accuracy of retrospective recall of opioid use over longer periods is unknown. Finally, although precautions were taken to ensure each Smart Cap access accurately recorded opioid intake, the technology cannot determine actual number of pills removed or consumed.

Overall, the results support retrospective self-reporting as a reasonably accurate method to assess prescribed opioid use in the home over a 2-week period. Further validation of patient-reported opioid use is recommended in additional surgical populations that include both sexes, more frequent preoperative opioid use, and for longer periods postoperatively and in chronic pain management populations.