Opioid Prescribing Rates in Nonmetropolitan and Metropolitan Counties Among Primary Care Providers Using an Electronic Health Record System — United States, 2014–2017

Macarena C. García, DrPH; Charles M. Heilig, PhD; Scott H. Lee, PhD; Mark Faul, PhD; Gery Guy, PhD; Michael F. Iademarco, MD; Katherine Hempstead, PhD; Dorrie Raymond, MA; Josh Gray, MBA

Disclosures

Morbidity and Mortality Weekly Report. 2019;68(2):25-30.

Abstract and Introduction

Introduction

Drug overdose is the leading cause of unintentional injury-associated death in the United States. Among 70,237 fatal drug overdoses in 2017, prescription opioids were involved in 17,029 (24.2%).^[1] Higher rates of opioid-related deaths have been recorded in nonmetropolitan (rural) areas.^[2] In 2017, 14 rural counties were among the 15 counties with the highest opioid prescribing rates.* Higher opioid prescribing rates put patients at risk for addiction and overdose.^[3] Using deidentified data from the Athenahealth electronic health record (EHR) system, opioid prescribing rates among 31,422 primary care providers^† in the United States were analyzed to evaluate trends from January 2014 to March 2017. This analysis assessed how prescribing practices varied among six urban-rural classification categories of counties, before and after the March 2016 release of CDC's Guideline for Prescribing Opioids for Chronic Pain (Guideline).^[4] Patients in noncore (the most rural) counties had an 87% higher chance of receiving an opioid prescription compared with persons in large central metropolitan counties during the study period. Across all six county groups, the odds of receiving an opioid prescription decreased significantly after March 2016. This decrease followed a flat trend during the preceding period in micropolitan and large central metropolitan county groups; in contrast, the decrease continued previous downward trends in the other four county groups. Data from EHRs can effectively supplement traditional surveillance methods for monitoring trends in opioid prescribing and other areas of public health importance, with minimal lag time under ideal conditions. As less densely populated areas appear to indicate both substantial progress in decreasing opioid prescribing and ongoing need for reduction, community health care practices and intervention programs must continue to be tailored to community characteristics.

Athenahealth is a commercial vendor and developer of cloud-based practice management and EHR systems for physician practices and hospitals. Approximately 100,000 health providers, serving about 86 million patients in the United States, use Athenahealth's applications. This retrospective study used deidentified Athenahealth EHR prescription data from 31,422 primary health care providers serving approximately 17 million patients. Patient-level data were aggregated by week over the 166 weeks from January 5, 2014, through March 11, 2017. For each week during which a patient had at least one Athenahealth record, that patient contributed one patient-week to this analysis. For each patient-week, it was noted whether primary care providers using Athenahealth's EHR system prescribed one or more opioids (Supplementary Table 1, https://stacks.cdc.gov/view/cdc/61743).^§ Percentage of patient-weeks during which an opioid prescription was written was considered equivalent to the percentage of patients receiving an opioid prescription during that time.

For comparisons over time, data were divided into three periods. Period 1 comprises 52 weeks from January 5, 2014, through January 3, 2015; period 2 includes the next 63 weeks, ending March 19, 2016; and period 3 covers the final 51 weeks, through March 11, 2017. The first cutpoint allows comparisons between the first and second years' data, and the second cutpoint supports comparisons before and after the publication of the CDC Guideline. For comparison by population density, data were stratified by providers' counties according to CDC's National Center for Health Statistics urban-rural classification scheme.^¶ From most to least densely populated, the six categories include large central metropolitan, large fringe metropolitan, medium metropolitan, small metropolitan, micropolitan, and noncore counties.

This analysis includes three components. First, the period-specific percentage of patients with opioid prescriptions was estimated empirically and with seasonal adjustment using logistic regression. Second, smooth temporal trends were statistically separated from annual seasonal components using locally weighted regression.^[5] Third, to quantify the period-specific annual rate of increase or decrease in prescribing rates, a second logistic regression model estimated the seasonally adjusted annual percent change (APC) in the odds of receiving an opioid prescription. Statistical software was used for all analyses; statistical tests and confidence intervals (CIs) are presented as simultaneous procedures adjusted for multiple comparisons.

Overall, 128,194,491 patient-weeks of data are included in the analysis; at least one opioid was prescribed during 8,810,237 (6.9%) of these patient-weeks, decreasing from 7.4% during period 1 to 6.4% during period 3 (Table 1) (Supplementary Table 2, https://stacks.cdc.gov/view/cdc/61744). Buprenorphine prescribed for pain and opioid use disorder treatment represented only 0.02% of all opioid prescriptions. By county classification, the overall percentage of patients with opioid prescriptions ranged from 5.2% in large central metropolitan counties to 9.6% in noncore counties during the study period. Patients in noncore counties had an 87% higher chance of receiving an opioid prescription than did patients in large central metropolitan areas during the study period.

The lowest period-specific percentages of patient-weeks with an opioid prescription occurred in large central metropolitan counties (5.0%–5.4%) (p<0.001, multiplicity-adjusted Wald tests), except during period 3, when percentages in large metropolitan counties (5.0%) were the same as those in large fringe metropolitan counties (5.0%) (Supplementary Table 2, https://stacks.cdc.gov/view/cdc/61744). In contrast, the highest period-specific percentages (9.0%–10.3%) were in noncore counties (p<0.02), except in period 3, when percentages in noncore counties (9.0%) were similar to those in micropolitan counties (9.1%). Across metropolitan and nonmetropolitan categories, all percentages of weeks with an opioid prescription during period 2 were significantly different from those in period 1, and percentages in period 3 differed significantly from those in period 2 (p<0.003).

Visual inspection of the prescribing trends by urban-rural status and by period revealed patterns in both the raw (Supplementary Figure 1, https://stacks.cdc.gov/view/cdc/61741) and seasonally adjusted (Supplementary Figure 2, https://stacks.cdc.gov/view/cdc/61742) data. During period 1, before release of the CDC Guideline, the odds of receiving an opioid prescription increased 6.4% per year in noncore counties (95% multiplicity-adjusted Wald CI = 2.1–10.8), and 9.7% per year in micropolitan counties (95% CI = 6.5–13.0) (Table 2) (Figure). During period 3, after release of the CDC Guideline, the odds of receiving an opioid prescription decreased significantly in all county groups. Comparing trends between periods, the APC increased in large central metropolitan counties in period 2 compared with period 1 (p<0.001) and decreased between periods 2 and 3 (p<0.001). In the other five urban-rural categories, the APC decreased in period 2 compared with period 1 (p<0.02); among these five groups, only micropolitan counties experienced a significant decrease in APC between periods 2 and 3 (p<0.001).

(Enlarge Image)

Figure.

Model-based trends in percentage of patient-weeks with at least one opioid prescription, by urban-rural category — Athenahealth, United States, January 2014–March 2017

1 2

Next Section

Abstract and Introduction
Discussion
References
Sidebar

* U.S. Opioid Prescribing Rate Maps. https://www.cdc.gov/drugoverdose/maps/rxrate-maps.html.
^† Primary care providers in an ambulatory setting; limited to family medicine, family practice, or general practice, or providers who have an internal medicine specialty with no subspecialty. Nurse practitioners and physician assistants are included among primary care providers.
^§ Short and long acting opioid drugs in this study included buprenorphine, butorphanol, codeine, dihydrocodeine, fentanyl, hydrocodone, hydromorphone, levorphanol, meperidine, methadone, morphine, naltrexone, nalbuphine, naloxone, oxycodone, oxymorphone, pentazocine, propoxyphene, tapentadol, and tramadol. The study does not count cough and cold medications containing opioids.
^¶ https://www.cdc.gov/nchs/data_access/urban_rural.htm; https://www.cdc.gov/nchs/data/series/sr_02/sr02_166.pdf.

Table 1. Number and percentage of patient-weeks with at least one opioid prescription — Athenahealth, United States, January 2014–March 2017
Urban-rural category*	No. of patient-weeks	No. receiving opioid prescription	Percentage receiving opioid prescription
Urban-rural category*	No. of patient-weeks	No. receiving opioid prescription	Overall	Period 1^†	Period 2^†	Period 3^†
Noncore	8,979,403	864,364	9.6	10.3	9.9	9.0
Micropolitan	16,342,824	1,532,747	9.4	9.4	9.6	9.1
Small metro	18,860,569	1,443,246	7.7	8.0	7.7	7.4
Medium metro	32,045,592	2,158,111	6.7	7.3	6.9	6.2
Large fringe metro	31,430,958	1,753,802	5.6	6.4	5.8	5.0
Large central metro	20,535,145	1,057,967	5.2	5.4	5.2	5.0
All counties	128,194,491	8,810,237	6.9	7.4	7.0	6.4

* National Center for Health Statistics urban-rural classification scheme for counties. https://www.cdc.gov/nchs/data_access/urban_rural.htm.
^† Period 1: January 5, 2014–January 3, 2015; period 2: January 4, 2015–March 19, 2016; period 3: March 20, 2016–March 11, 2017. Period-specific percentages are based on raw counts rather than statistical models.

Table 2. Annual percent change (APC) in odds of receiving at least one opioid prescription — Athenahealth, United States, January 2014–March 2017
Urban-rural category*	Period 1^†	Period 2^†	Period 3^†	p-value (direction of change)^§
Urban-rural category*	APC (95% CI)	APC (95% CI)	APC (95% CI)	Period 1 versus period 2	Period 2 versus period 3
Noncore	6.4 (2.1 to 10.8)^¶	-10.1 (-12.2 to -8.0)^¶	-7.5 (-10.7 to -4.2)^¶	<0.001 (decrease)	0.713 (—)
Micropolitan	9.7 (6.5 to 13.0)^¶	-0.8 (-2.6 to 0.9)	-13.3 (-15.6 to -10.9)^¶	<0.001 (decrease)	<0.001 (decrease)
Small metro	0.2 (-2.8 to 3.2)	-4.5 (-6.2 to -2.7)^¶	-5.8 (-8.4 to -3.2)^¶	0.013 (decrease)	0.977 (—)
Medium metro	-2.5 (-4.8 to -0.1)**	-8.7 (-10.1 to -7.4)^¶	-9.2 (-11.2 to -7.2)^¶	<0.001 (decrease)	0.999 (—)
Large fringe metro	-2.0 (-4.7 to 0.8)	-14.9 (-16.2 to -13.5)^¶	-13.1 (-15.1 to -10.9)^¶	<0.001 (decrease)	0.616 (—)
Large central metro	-9.9 (-13.2 to -6.4)^¶	1.8 (-0.3 to 3.9)	-11.7 (-14.3 to -8.9)^¶	<0.001 (increase)	<0.001 (decrease)
All counties	-1.4 (-10.6 to 8.8)	-8.2 (-13.3 to -2.7)^††	-10.4 (-17.9 to -2.2) ^††	0.371 (—)	0.856 (—)

Abbreviation: CI = confidence interval.
* National Center for Health Statistics urban-rural classification scheme for counties. https://www.cdc.gov/nchs/data_access/urban_rural.htm.
^† Period 1: January 5, 2014–January 3, 2015; period 2: January 4, 2015–March 19, 2016; period 3: March 20, 2016–March 11, 2017.^§ p-values from multiplicity-adjusted Wald tests; (—) indicates a nonsignificant difference (p>0.05) between APCs in adjacent periods.
^¶ p<0.001.
** p<0.05.
^†† p<0.01.