Open access

Research Article

13 May 2020

After-Hours Incentives and Emergency Department Visits: Evidence from Ontario

Authors: Rose Anne Devlin, Koffi Ahoto Kpelitse, Lihua Li, Nirav Mehta, and Sisira SarmaAuthor information and affiliations

Publication: Canadian Public Policy

Volume 46, Number 2

https://doi.org/10.3138/cpp.2019-046

PDF EPUB

Abstract

One important component of primary care reform in Ontario is to incentivize physicians to work after hours to improve access to core primary care services and potentially reduce visits to hospital emergency departments (EDs). Empirically, evidence on this link is ambiguous. We suggest reasons for this ambiguity and then harness rich administrative data from Ontario to carefully investigate whether and why after-hours incentives affect ED usage. The data cover physicians’ office visits and ED visits from 2003 to 2007, a period with exogenous changes in after-hours incentives. We find strong evidence that less urgent ED visits are reduced as a result of these incentives.

Résumé

Un aspect important de la réforme des soins de santé primaires en Ontario est d’inciter les médecins à effectuer des heures supplémentaires pour améliorer l’accès aux principaux services de soins primaires et pour réduire le nombre des visites aux urgences dans les hôpitaux. Les données empiriques relatives à ce lien sont ambiguës. Les auteurs tentent d’expliquer cette ambiguïté et se servent de riches données administratives ontariennes pour analyser attentivement si les mesures d’encouragement aux heures supplémentaires influent sur le recours aux services d’urgence et pourquoi. Les données étudiées portent sur les visites aux cabinets des médecins et les visites aux urgences de 2003 à 2007, période marquée par des changements exogènes dans les mesures d’encouragement aux heures supplémentaires. Selon les auteurs, tout porte à croire que ces mesures d’encouragement contribuent à diminuer les visites moins urgentes en salles d’urgence.

Introduction

Avoidable, or less urgent, emergency department (ED) visits are commonplace in developed countries (Carret, Fassa, and Domingues 2009). About 39 percent of Canadians reported that their ED visits could have been avoided if primary care were available (Schoen et al. 2005). Aside from contributing to overcrowding and delaying care for patients in urgent need, the use of EDs for less urgent health problems contributes to higher health care costs (Campbell et al. 2005; Mehrotra et al. 2009; Thygeson et al. 2008) and lower continuity of care, which adversely affects health outcomes, especially for patients with chronic conditions (Dunnion and Kelly 2005; Stiell et al. 2003; Vinker et al. 2004).

Primary care is publicly funded in all jurisdictions in Canada with no direct cost to the patient for physician and hospital visits. Although no direct financial cost is borne, ED visits entail long waits before a physician is seen, especially for less urgent cases. All else being equal, patients would likely prefer to be treated by their family doctor for less urgent health problems.

Reforms to primary care were introduced in the early 2000s across various jurisdictions in Canada (Gray et al. 2015; Health Canada 2007; Sweetman and Buckley 2014). Common across these reforms was a move from traditional fee-for-service (FFS) remuneration toward pay-for-performance incentives for preventive care and chronic disease management (Hutchison et al. 2011). Several new types of non-FFS primary care delivery models that feature these financial incentives have been introduced in Canada’s most populous province, Ontario, since 2004 (Hutchison et al. 2011). By 2010, more than two-thirds of Ontario’s family physicians had joined one of these models, with Family Health Organizations (FHOs) and Family Health Groups being the two most popular choices (Henry et al. 2012).

One of the goals of these new models is to reduce ED visits by increasing access to primary health care services outside of regular working hours. Each physician practicing in these new models is required to provide a minimum of one three-hour session per week either weeknights after 5:00 p.m. or on weekends or statutory holidays. In return, these physicians receive an after-hours premium, which was initially 10 percent when first introduced in 2003 and then increased to 15 percent in April 2005, 20 percent in April 2006, and 30 percent in September 2011 (Sweetman and Buckley 2014). The main goal of this article is to determine whether this policy was successful in reducing ED utilization.

Pay-for-performance schemes have been studied in several contexts in Ontario (Kantarevic and Kralj 2013; Li et al. 2014). A handful of studies examine the link between improved after-hours access to primary care and ED visits outside of Canada, with mixed findings. This literature includes the implementation of an after-hours clinic or cooperative (Buckley, Curtis, and McGirr 2010; Pickin et al. 2004), the extension of primary care practice opening hours (Dolton and Pathania 2016; Harris, Patel, and Bowen 2011; Lippi Bruni, Mammi, and Ugolini 2016; Lowe et al. 2005), the reorganization of after-hours care (van Uden and Crebolder 2004; van Uden et al. 2005), and after-hours financial incentives (Franco, Mitchell, and Buzon 1997; Piehl, Clemens, and Joines 2000). This literature guides our choice of variables in the analyses.

At least two reasons explain why increased after-hours services may not unambiguously reduce ED visits: first, ED visits would fall only if patients have conditions that are otherwise treatable by primary care physicians; no amount of after-hours care would reduce visits to the ED for patients with the most urgent needs. Second, although ED visits may fall because of increased after-hours services, they may rise if regular-hours services are reduced.

We use a rich longitudinal data set that allows us to control otherwise unobserved heterogeneity and to exploit exogenous variation in the strength of after-hours incentives, which helps to establish the impact of these incentives on ED visits. The large number of physicians in this data set also allows us to estimate this impact by different subgroups of physicians, such as those with sicker patients. As expected, regular- and after-hours services move in opposite directions in response to stronger after-hours incentives. We find that after-hours services reduce less urgent ED visits. Most less urgent ED reductions come from practices with below-median co-morbidity, suggesting that after-hours incentives reduce ED visits by healthier patients, allowing more time for more urgent visits.

Data and Variables

Several administrative databases held at the Institute for Clinical Evaluative Sciences provide the data for this study. The Physician Database contains characteristics of primary care physicians; the Corporate Provider Database provides physicians’ model type, effective date of eligibility for billing under the Ontario Health Insurance Plan (OHIP; the public insurer/payer), and physician group size. The Client Agency Program Enrollment Database allows us to match physicians with enrolled patients. If a physician was affiliated with more than one practice type, the most recent one joined was selected. Only physicians who put in a claim for after-hours incentives to the public insurer (OHIP) are included in our sample.¹ The billing codes eligible for the after-hours premium correspond to fairly basic services that most physicians provide, such as minor assessments, primary mental health care, counseling, and annual physical examinations. Knowing this, we expected an appreciable effect of premium changes on ED utilization. We exclude part-time physicians, defined as having fewer than 500 patients or 500 visits in any given year. We focus on data spanning 2003–2007, with 1,321 unique physicians and 6,605 physician-year observations comprising the balanced panel we use for our analysis.²

Patient visits to a physician were identified through OHIP billing claims. For each physician, total annual office visits were derived as the sum of patient visits.³ The total number of annual office visits minus the total number of annual after-hours visits defines regular visits for each physician. Group size sums up the number of primary care physicians with the same group number.

ED visits come from the National Ambulatory Care Reporting System, which classifies them into urgent and non-urgent on the basis of the Canadian Triage and Acuity Scale (CTAS) (Beveridge, Clarke, and Janes 1999). A triage level of 1 (resuscitation), 2 (emergent), or 3 (urgent) is urgent, and a level of 4 (less urgent/semi-urgent) or 5 (non-urgent) was not; we aggregate these into three mutually exclusive categories, where Group 1 (very urgent) contains ED visits with a CTAS score of 1 or 2, Group 2 (urgent) contains ED visits with a CTAS score of 3, and Group 3 (less urgent) contains ED visits with a CTAS score of 4 or 5.⁴ The Aggregated Diagnosis Group (ADG) reflecting the health status of each patient is based on the patient’s diagnosis codes from the hospital Discharge Abstract Database and OHIP, using the Johns Hopkins Adjusted Clinical Group case-mix adjustment system. There are 32 diagnosis groups, which we sum so that each patient has a score between 1 and 32. We calculate the average ADG for each physician’s patients.

Ontario’s health registry database (the Registered Persons Database) provides patients’ age, sex, and postal codes, which we use to obtain deprivation and rurality indices around the patient’s residence (census dissemination area). The deprivation index is organized into quintiles, where 1 is least marginalized and 5 is most marginalized—our measure is the percentage of physician’s patients from the fourth and fifth quintiles (i.e., the most deprived areas)—and individuals with a rurality index of 40 or higher are considered to reside in rural areas (Kralj 2000; Matheson et al. 2012).⁵

Table 1 reports descriptive statistics of regular-hours, after-hours, and ED utilization by year. Between 2003 and 2007, the number of regular-hours visits per 1,000 patients decreased; after-hours visits per 1,000 patients increased and then fell near the end. Correspondingly, total costs per 1,000 patients decreased substantially, even though after-hours costs almost doubled.⁶ During the same period, the number of very urgent ED visits increased, urgent ED visits increased sharply initially and then stayed roughly constant, and less urgent ED visits also increased sharply and then gradually decreased. Table 2 presents statistics pooled over the sample period and shows that after-hours visits make up about 11 percent of physicians’ total visits on average.

Empirical Framework

The net effect of increasing the after-hours premium on ED visits is ambiguous. We start by estimating the regression⁷

μ_{i t} = ρ_{μ, π} π_{t} + Z_{i t}^{'} β_{μ, π} + ϵ_{i t, μ, π},

(1)

where µ_it can be total, less urgent, urgent, or very urgent ED visits per 1,000 patients for physician i in year t; π_t is the after-hours premium in year t; Z_it includes a time trend, physician’s age, physician’s age squared, proportion of female physicians and foreign graduates in the physician’s practice, group size, average age of patients, average ADG score of patients, proportion of patients living in deprived areas, and proportion of patients living in rural areas; and

ϵ_{i t, μ, π}

represents the error term. We estimate the regression using both Ordinary Least Squares (OLS) and physician fixed effects (FE), which means the error term may include a FE component for the physician. However, the net effect of increasing the after-hours premium on overall costs may be ambiguous, even if less urgent ED visits decrease as a result of the after-hours premium increase. We thus re-estimate the model using total costs as the dependent variable.

To examine how physician behaviour is affected by incentivizing after-hours access, we estimate a model of services provided by physician i during year t (using OLS and physician FE):

x_{i t} = ρ_{x, π} π_{t} + Z_{i t}^{'} β_{x, π} + ϵ_{i t, x, π},

(2)

where x_it can be either regular- or after-hours services (analyzed separately), measured by either visits per 1,000 patients or (deflated) costs per 1,000 patients (i.e., the value of services in 2004 prices); π_t is the premium level in year t; Z_it contains the variables from Equation (1); and

ϵ_{i t, x, π}

represents the error term, which may include a FE component for the physician. Because FFS physicians do not receive after-hours premiums, we focus our analysis on physicians who have switched into a scheme that incentivizes after-hours services (recall that we restrict our analysis to go through 2007, before the large transition to FHO occurred).

Next, we estimate how ED visits vary with respect to the value of after-hours services in 2004 prices, using the regression model (estimated using OLS and physician FE)

μ_{i t} = ρ_{μ, x} x_{i t} + Z_{i t}^{'} β_{μ, x} + ϵ_{i t, μ, x},

(3)

where µ_it can be total, less urgent, urgent, or very urgent ED visits per 1,000 patients for physician i in year t; x_it represents the value of after-hours services in 2004 prices; Z_it contains the variables from Equation (1); and

ϵ_{i t, μ, x}

represents the error term, which may include an FE component for the physician. One might wonder whether physicians would engage in “gaming” in response to the financial incentives. They could, for instance, encourage their patients to come during after hours in order to gain the premium. If all changes in after-hours services were driven by gaming behaviour there would be no change in the provision of primary care and, thus, no effect on ED visits. Our empirical approach takes into account this potential for gaming, as we estimate the net effect of premium increases on ED utilization. Evidence of an effect of after-hours premium increases on ED visits would suggest such gaming is not the dominant force at play. From a practical perspective, however, the regulatory framework affords very limited scope for gaming. Physicians are required to post their after-hours availability to patients, and the incentives are applicable only to enrolled patients seen during posted after-hours sessions. Physicians are free to provide services to non-enrolled patients but these are not eligible for after-hours incentives.

Table 1: Sample Means, by Year (2003–2007)

Variable	2003	2004	2005	2006	2007
Premium, %	0.10	0.10	0.15	0.20	0.20
No. of observations	1,321	1,321	1,321	1,321	1,321
No. of regular-hours visits^a	4,129.85	3,096.45	2,968.89	2,933.39	2,751.81
No. of after-hours visits^a	224.94	402.10	408.39	379.91	350.26
Total costs (defl ated), 2000 $^a	173,668.80	142,387.30	140,541.00	134,611.60	125,896.30
After-hours costs (defl ated), 2000 $^a	6,295.60	11,251.11	11,860.40	11,073.03	10,228.15
No. of ED visits^a
Very urgent	35.67	43.40	47.79	50.37	54.65
Urgent	138.58	147.07	151.62	152.62	152.89
Less urgent	138.58	183.78	177.37	176.03	173.65

Notes: Very urgent, urgent, and less urgent ED visits, respectively, correspond to CTAS Groups 1, 2, and 3. CTAS = Canadian Triage and Acuity Scale; ED = emergency department.

Per 1,000 patients.

Source: Authors’ calculations.

Table 2: Sample Means and Standard Deviations, Pooled over All Years

Variable	Mean (SD)
Age, y	48.96 (8.61)
Female (1 = female)	0.32 (0.46)
IMG (1 = IMG)	0.08 (0.26)
Group size, no. of patients	23.18 (28.33)
Roster size, no. of patients	1,773.75 (666.15)
Avg. age, y	38.47 (5.29)
Avg. ADG (ADG units = 1–32)	3.18 (0.38)
Avg. deprived, %	23.12 (13.02)
Avg. rural, %	12.50 (22.20)
Premium, %	0.12 (0.05)
Visits, no.^a
Regular hours	3,176.08 (1,006.39)
A fter hours	353.12 (319.02)
ED visits, no.^a
Very urgent	46.38 (21.83)
Urgent	148.56 (49.30)
Less urgent	169.88 (119.86)
Costs (defl ated), $^a
After hours	10,141.66 (9,317.56)
Total	143,421.00 (47,742.82)
ED costs, $^a
Very urgent	16,876.07 (8,254.99)
Urgent	35,443.63 (11,456.86)
Less urgent	25,882.34 (18,176.57)
Total	78,210.01 (27,569.96)
No. of obs.	6,605

Notes: Very urgent, urgent, and less urgent ED categories, respectively, correspond to CTAS Groups 1, 2, and 3. ADG = Aggregated Diagnosis Group; Avg. = average; CTAS = Canadian Triage and Acuity Groups; IMG = international medical graduates; obs. = observations.

Per 1,000 patients.

Source: Authors’ calculations.

Empirical Findings

Effect of After-Hours Incentives on ED Utilization

Table 3 provides the estimated coefficients from the OLS and FE models of ED visits (total, and split by urgency group) on the after-hours premium as well as physician and practice characteristics (equation (1)). From the OLS results in the left panel, the estimated coefficients on the after-hours premium (premium) are negative across the board. Not only do total ED visits fall, but so too do the visits of all groups (with less urgent visits being statistically insignificant). The last result is surprising insofar as one would expect that less urgent visits would decrease after the introduction/strengthening of after-hours incentives for regular physicians. The OLS procedure does not control for time-invariant physician heterogeneity which affects the reliability of these estimates. Indeed, when we take account of physician level FE, these results change quite significantly. From the right-hand side of Table 3, we find that the after-hours premium exerts a negative and statistically significant impact for the less urgent group (specification (6)) and a positive and statistically significant impact on more urgent ED visits (specifications (8) and (7)). Together, these findings support the idea that the after-hours premium encouraged those patients with less urgent conditions to seek care at their physician’s office, thus reducing their crowding out of more urgent ED visits. We also see that female physicians tend to have fewer ED visits, as do international medical graduates (IMG) and those physicians who practice in larger practices (group size). Physicians with younger patients (avg. age), fewer deprived patients (avg. deprived), and fewer rural patients (avg. rural) also have fewer ED visits. Physicians with higher mean patient ADGs (avg. ADG) have more very urgent ED visits and fewer less urgent ED visits.

Effect of After-Hours Incentives on Services

If increasing the after-hours premium reduces ED utilization by changing physician behaviour, this should show up in the data. Office visits are one measure of physician behaviour but may not be ideal because the amount of services provided per visit may change if the after-hours premium changed. The rich data used in this study enable us to use two measures of physician after-hours services: visits and the value of services provided. Value of services provided are a good measure of services rendered if service prices capture input amounts. We construct price-adjusted values, “after-hours values” and “regular values”, by correcting for the change in the prices of services (in 2004 values), excluding the after-hours bonuses (which were, e.g., 10 percent in 2004). We use these deflated values as our primary measure of services in our empirical analysis.

Table 4 presents the FE results from regressing measures of services on the after-hours premium (Equation [2]). Specification (1) regresses after-hours visits on the premium and finds a positive and significant estimated coefficient: a higher premium is correlated with more after-hours visits. Specification (2) uses the (deflated) value of after-hours services—that is, the cost, adjusted for changes in the premium and in the prices of services—as the measure of services rendered and finds a strong positive relationship between the premium amount and services. Specification (3) examines how the after-hours premium affects regular visits and reveals the expected negative, significant effect. Specification (4) examines how the (deflated) value of all services co-varies with the premium and finds a negative, significant effect. Because total costs include regular- and after-hours costs, this estimate implies that the regular-costs measure of services provided during regular hours must have decreased. In short, both measures of services provided after hours increase in response to increases in the premium, and both measures of services provided during regular hours decrease in response to premium increases.

Effect of After-Hours Services on ED Utilization

The FE estimates of the impact of the after-hours premium on ED visits indicate that they fell for the less urgent group (Table 3), whereas after-hours physician services, not surprisingly, increased (Table 4). Table 5 summarizes the relationship between after-hours costs and ED utilization, using OLS (first row) and FE (second row) estimates of Equation (3). The OLS results are presented for comparison purposes, but we focus on the FE ones. The FE estimates show that, although increases in after-hours services do not decrease overall ED visits (Specification [1]), they do significantly decrease the least urgent ED visits (Specification [2]); they are also associated with small (but statistically significant) increases in urgent ED visits (specification [3]) but are not associated with any change in very urgent ED visits (specification [4]).

The FE models control for time-invariant physician heterogeneity and provide evidence that increases in after-hours services reduce less urgent ED visits. We can provide even stronger evidence supporting the link between after-hours services and ED visits by exploiting variation in the after-hours premium over time as an instrument and using a FE instrumental variables estimation approach. Because changes in the after-hours premium over time are exogenous,⁸ the premium is arguably a valid instrument in the ED utilization equations. Given the higher wage rate faced by physicians during after-hours sessions as well as regulations governing the minimum number of these sessions, we expect the after-hours premium to be highly correlated with after-hours services. This is confirmed by the large F statistics in the first-stage regression (see Table 6). Moreover, there is no a priori reason to expect the after-hours premium awarded family physicians to affect ED utilization other than indirectly through the increased provision of incentivized services during after hours.

Specification (1) of Table 6 shows that there is no significant estimated impact of after-hours services on total ED visits. However, Specification (2) shows that there is a significant reduction in the least urgent ED visits stemming from increases in after-hours services. Estimates of the effect on urgent and very urgent ED visits are positive and statistically significant, yet smaller in magnitude than the effect on less urgent ED visits. These results corroborate those of the FE models; namely, increasing the after-hours premium increases physicians’ provision of after-hours services, which leads to a reduction in less urgent ED visits. The advantage of the instrumental variables estimates is that they isolate physician behaviour as the channel affecting ED utilization.

Table 3: Change in ED Visits as a Function of After-Hours Premium and Other Covariates

Variable	OLS								FE
	Total Visits (1)		Less Urgent (2)		Urgent (3)		Very Urgent (4)		Total Visits (5)		Less Urgent (6)		Urgent (7)		Very Urgent (8)
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Premium	−126.261	33.860	−1.353	27.732	−33.918	16.837	−16.591	8.732	32.307	17.137	−56.890	21.046	43.297	6.648	11.839	4.239
Year	4.120	1.437	1.786	1.166	2.990	0.698	4.472	0.370	−10.556	3.310	0.084	3.838	−0.087	1.277	3.391	0.771
Age	−7.493	2.932	−1.919	2.011	−2.899	1.108	−0.518	0.455
Age²	0.055	0.029	0.008	0.020	0.022	0.011	0.005	0.005	0.135	0.028	−0.008	0.036	0.017	0.012	0.004	0.007
Female	−13.213	6.584	4.550	4.263	−6.057	2.449	−7.684	0.960
IMG	−38.437	9.915	−21.445	6.597	−13.093	3.698	−7.419	1.677
Group size	−0.331	0.060	−0.177	0.039	−0.184	0.025	−0.025	0.011	0.048	0.026	−0.079	0.028	−0.013	0.011	0.010	0.006
Avg. age	5.207	0.705	2.195	0.469	2.312	0.245	1.058	0.109	1.208	1.653	0.987	1.501	0.464	0.450	0.547	0.210
Avg. ADG	−12.750	8.329	−20.765	5.652	0.950	3.073	6.276	1.317	8.862	8.302	8.728	7.747	7.897	2.486	1.964	1.461
Avg. deprived	3.648	0.260	2.662	0.267	1.300	0.096	0.447	0.046	0.382	0.159	5.103	0.566	0.171	0.059	0.096	0.027
Avg. rural	3.979	0.303	3.061	0.249	0.303	0.064	−0.148	0.018	3.580	1.083	−0.177	0.504	0.298	0.154	−0.046	0.107
Constant	−7918.8	2879.5	−3449.7	2339.4	−5874.5	1397.5	−8969.0	739.6	21072.0	6545.4	−150.7	7588.3	225.3	2530.7	−6791.7	1528.8
No. of obs.	6,605		6,605		6,605		6,605		6,605		6,605		6,605		6,605

Notes: All regressions include the full set of control variables, Z_it. FE are at the physician level. SEs are clustered at the physician level; there are 1,321 clusters. ADG = Aggregated Diagnosis Group; Avg. = average; Coef. = coeffi cient; ED = emergency department; FE = fi xed effects; IMG = international medical graduates; obs. = observations; OLS = Ordinary Least Squares.

Source: Authors’ calculations.

Table 4: FE Regressions of Regular- and After-Hours Services on After-Hours Premium and Other Characteristics

Variable	After-Hours Visits (1)		After-Hours Costs (2)		Regular Visits (3)		Total Costs (4)
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Premium, %	436.649	69.423	13,166.930	2,048.579	−2,223.516	244.539	−58,298.540	10,185.040

Notes: All regressions include the full set of control variables, Z_it. Fixed effects are at the physician level. Standard errors are clustered at the physician level; there are 6,605 observations and 1,321 clusters. Full results are provided in Online Appendix Table A.5. Coef. = coeffi cient; FE = fixed effects.

Source: Authors’ calculations.

Table 5: Change in ED Visits with Respect to After-Hours Services

Variable	Total visits (1)		Less urgent (2)		Urgent (3)		Very urgent (4)
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
OLS	−0.00280	0.00027	−0.00172	0.00020	−0.00082	0.00010	−0.00006	0.00006
FE	0.00019	0.00015	−0.00127	0.00025	0.00018	0.00006	0.00006	0.00004

Notes: All regressions include the full set of control variables, Z_it. After-hours services are measured using defl ated after-hours costs. FE are at physician level. SEs are clustered at physician level; there are 6,605 observations and 1,321 clusters. Coef. = coeffi cient; ED = emergency department; FE = fi xed effects; OLS = ordinary least squares.

Source: Authors’ calculations.

Table 6: IV Estimate of Effect of After-Hours Services on ED Visits

Variable	Total Visits (1)		Less Urgent (2)		Urgent (3)		Very Urgent (4)
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
After-hours costs	0.00246	0.00134	−0.00432	0.00165	0.00329	0.00069	0.00090	0.00036
F (1, 1320)	41.20

Notes: The Stata command xtivreg2 was used to estimate these regressions. All regressions include the full set of control variables, Z_it. Afterhours services are measured using defl ated after-hours costs. SEs are clustered at physician level; there are 6,605 observations and 1,321 clusters. Coef. = coeffi cient; ED = emergency department; IV = instrumental variable.

Source: Authors’ calculations.

Up till now, we see that increasing the after-hours premium reduces ED visits for less urgent patients by way of increasing physicians’ after-hours services. Now we want to determine whether this affects overall health system costs. Using values for ED visits (2002 dollars) from the Ministry of Health,⁹ we estimate

$ μ_{i t} = ρ_{$ μ, x} $ x_{i t} + Z_{i t}^{'} β_{$ μ, x} + ϵ_{i t, $ μ, x^{'}}

(4)

where $µ_it is the cost of ED visits per 1,000 patients and $x_it is the cost of the after-hours incentive per 1,000 patients. Note that the cost of the after-hours incentive used in this specification includes the increase in premium. Here, we are primarily interested in the estimate of ρ_$µ,x, the change in ED costs with respect to after-hours costs.

Table 7 presents FE instrumental variables estimates of after-hours services on ED costs. The increase in after-hours services (resulting from the higher after-hours premium) results in reductions in the least urgent ED costs (Specification [2]). By dividing the estimated ED cost savings by the estimated reduction in less urgent ED visits,

\frac{- $ 0.6442 per 1, 000 patients per physician per year}{- 0.00432 visits per 1, 000 per 1, 000 patients per physician per year}

we see a cost savings of about $149 per reduced ED visit. This calculation exploits changes in costs and visits induced by variations in the after-hours premium, and as such it represents the mean costs of treating inframarginal patients at the less urgent ED. The increase in costs resulting from more urgent care visits is of a much smaller magnitude than the cost savings found for less urgent ED visits. Overall, our results suggest that the after-hours policy contributed to a reduction in ED costs.

Table 7: IV Estimate of Effect of After-Hours Services on ED Costs

Variable	Total Cost (1)		Less-Urgent Cost (2)		Urgent Cost (3)		Very Urgent Cost (4)
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
After-hours costs	−0.1979	0.2452	−0.6442	0.1908	0.2258	0.1202	0.2196	0.1320
F (1, 1320)	41.20

Notes: Very urgent, urgent, and less urgent ED categories correspond, respectively, to CTAS Groups 1, 2, and 3. The Stata command xtivreg2 was used to estimate these regressions. All regressions include the full set of control variables, Z_it. After-hours services are measured using defl ated after-hours costs. SEs are clustered at physician level; there are 6,605 observations and 1,321 clusters. Coef. = coeffi cient; CTAS = Canadian Triage and Acuity Scale; ED = emergency department; IV = instrumental variable.

Source: Authors’ calculations.

Finally, it is likely that reductions in ED visits are not uniformly distributed over all practices. In particular, in practices with healthier patients, increased after-hours services may allow patients with less urgent conditions to be treated by their physicians rather than at the ED, as opposed to practices with high-morbidity patients who may need to go to the ED irrespective of the availability of after-hours sessions. To examine this possibility, we re-run our prior FE regressions for two subsamples of practices: those with below-median average patient ADG and those with above-median average patient ADG.

The first row of Table 8 reports the change in ED visits with respect to after-hours services (measured by after-hours costs), split by whether practices have low and high mean ADGs. The results from Specifications (3) and (4) indicate a significant negative relationship with respect to the value of after-hours services and less urgent ED visits; the estimated relationship for urgent ED visits is positive, but much smaller in magnitude. In the second row, we find that the response of less urgent ED costs with respect to after-hours services is –0.109 for practices with below-median mean ADG, whereas it is not significantly different from zero at practices with above-median mean ADG. The finding that changes in ED costs are driven by physicians with relatively healthy (and, therefore, easier to treat) patients is consistent with the idea that premium changes affect ED utilization via physician behaviour.

Concluding Remarks

Our results show that the after-hours incentive in Ontario’s primary care setting resulted in more patients with less urgent needs seeing their family physicians, reducing less urgent ED visits. Our article has several strengths. We provide empirical evidence using novel health administrative data from Ontario that after-hours incentives reduce ED visits. Moreover, we are able to investigate the impact of after-hours incentives on ED utilization at the intensive margin. We uncover evidence that after-hours incentives reduce less urgent ED utilization, stemming largely from practices with healthier patients rather than from those with sicker ones.

We noted earlier that, although we have data for 2003–2013, we focus our analysis on 2003–2007. In addition to an increase in the after-hours premium from 20 percent to 30 percent, several additional changes occurred in the institutional environment from 2008 to 2013, including the introduction of additional pay-for-performance incentives and the switch of many physicians to FHO and team settings, which was found to reduce services in FHO baskets (Zhang and Sweetman 2018). We conducted a robustness analysis covering this later period and found several qualitatively similar results, which we present in the Online Appendix. Briefly, similar to the 2003–2007 period, online Appendix Table A.1 shows that increases in the premium are associated with statistically significant reductions in less urgent ED visits (specifications [2] and [6]). Online Appendix Table A.2 shows that there are significant reductions in regular visits and costs (specifications [3] and [4]), although Specifications (1) and (2) do not show significant increases in after-hours services, as measured by either visits or costs. The within-physician (FE) estimates presented in the second row of online Appendix Table A.3 show that increases in after-hours services reduce less urgent ED visits (Specification [2]), similar to the results from 2003–2007. Finally, similar to our results from 2003–2007, online Appendix Table A.4 shows significant reductions in ED costs associated with less urgent visits (specification [2]).

One interesting direction for future work concerns the after-hours premium itself. For example, a constant after-hours premium may not be the most cost effective, and one that differs with disease severity may lead to further health system savings. Examining the optimal non-linear incentive structure would be fruitful. Another point for future research is whether after-hours incentives work differently in retrospective and perspective payment systems. Finally, it would be useful and pertinent to examine the extent to which these reforms resulted in more individuals finding a regular family doctor—potentially reducing their reliance on ED visits.

Like other studies using administrative data, we are limited as to the variables available for empirical analyses. The lack of socio-economic information on physicians and their patients, including information on family income, constrains the work. Such data would enhance future research in this area.

Table 8: Changes in ED Visits and Costs with Respect to After-Hours Services

Variable	Total				Less Urgent				Urgent				Very Urgent
	Low Average ADG (1)		Low Average ADG (2)		Low Average ADG (3)		High Average ADG (4)		Low Average ADG (5)		Low Average ADG (6)		Low Average ADG (7)		Low Average ADG (8)
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
ED Visits	–0.00006	0.00035	0.00019	0.00015	–0.00112	0.00052	–0.00114	0.00031	0.00049	0.00014	0.00006	0.00007	0.00009	0.00006	–0.000006	0.000047
ED Costs	–0.00299	0.06217	0.02277	0.03462	–0.10850	0.04067	0.01807	0.01553	0.08699	0.03354	0.00768	0.02129	0.01833	0.02239	–0.002996	0.016627

Notes: Very urgent, urgent, and less urgent ED categories correspond, respectively, to CTAS Groups 1, 2, and 3. “Low Average ADG” and “High Average ADG,” respectively, refer to practices in which mean patient ADG is below and above the sample median mean practice ADG. All regressions include the full set of control variables, Z_it. After-hours services are measured using deflated after-hours costs. FE are at the physician level. Standard errors are clustered at physician level; there are 6,605 observations and 1,321 clusters. ADG = Aggregated Diagnosis Group; Coef. = coefficient; CTAS = Canadian Triage and Acuity Scale; ED = emergency department; FE = fixed effects.

Source: Authors’ calculations.

Acknowledgements

An earlier version of this article was presented at the 49th Annual Conference of the Canadian Economics Association, the 16th Annual Canadian Health Economists’ Study Group meeting, and the 2017 International Health Economics Association Congress, and we are grateful for various comments of participants and discussants. Two anonymous referees provided many useful suggestions. We thank Alex Kopp, Sue Schultz, Rick Glazier, Salimah Shariff, and Amit Garg at the Institute for Clinical Evaluative Sciences (ICES) for help in moving forward with the ICES data. Funding for this research by the Canadian Institutes of Health Research operating grant (MOP–130354) is gratefully acknowledged. This study was undertaken at the ICES Western site. ICES is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). Core funding for ICES Western is provided by the Academic Medical Organization of Southwestern Ontario (AMOSO), the Schulich School of Medicine and Dentistry (SSMD), Western University, and the Lawson Health Research Institute (LHRI). The opinions, results, and conclusions are those of the authors and are independent from the funding sources. No endorsement by ICES, AMOSO, SSMD, LHRI, Canadian Institutes of Health Research, or the MOHLTC is intended or should be inferred.

Footnotes

The OHIP billing codes used to compute after-hours costs are A001, A003, A004, A007, A008, A888, K005, K013, K017, K033, K030, Q050, K130, K131, and K132; the after-hours premium codes are Q12 and Q16.

We have data covering 2003–2013. However, beginning in 2008, a large number of family physicians switched to the FHO model, a capitation-based payment system found to change physician practice patterns (Zhang and Sweetman 2018). Because these changes may confound our analysis of the impact of after-hours incentives on ED visits, we restricted our analysis to 2003–2007. Moreover, using a balanced panel eliminates potential attrition bias stemming from the retirement of physicians.

Patient visits are defined as the combination of billing codes and the serve date in OHIP (i.e., the date the service or services were provided to patients). Multiple billing codes on the same date are defined as a one visit, but two billing codes on two separate dates are defined as two visits.

Unfortunately, the sample sizes for those at CTAS Levels 1 and 5 were not large enough to permit us to use all five categories in our analyses.

The (material) deprivation index is a composite score based on the proportion of the population in the census area aged 25 years or older without a certificate, diploma, or degree; the proportion of single-parent families; the proportion receiving government transfer payments, the proportion of those aged 15 years or older who are unemployed, the proportion considered low income, and the proportion living in homes in need of major repair; ethnic concentration is a composite score based on neighbourhood-level proportions, including the proportion who are recent immigrants (within 5 years) and the proportion of those who identify as self-minorities (Matheson, Moloney, and Van Ingen 2018).

Note that by “after-hours costs” we refer to the value of after-hours services, excluding the after-hours premium.

We adopt the notational convention of subscripting regression coefficients with the dependent variable first and then the primary regressor of interest.

Even if the policy to increase after-hours premiums were made in response to an undesirably large number of ED visits, it is exogenous to the actions of any given physician.

Specifically, ED costs were derived by multiplying the resource intensity weight available from the National Ambulatory Care Reporting database with the average cost case by the cost per weighted case. This is the standard procedure commonly used to calculate costs at the population level (Wodchis et al. 2013).

Supplementary Material

Appendix (cpp.2019-046_appendix1.pdf)

Download
403.63 KB

References

Beveridge, R., B. Clarke, and L. Janes. 1999. “Canadian Emergency Department Triage and Acuity Scale Implementation Guidelines.” Canadian Journal of Emergency Medicine 1(Supplement):S2–S20.

Format	RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks
Direct import

Abstract

Abstract

Résumé

Introduction

Data and Variables

Empirical Framework

Empirical Findings

Effect of After-Hours Incentives on ED Utilization

Effect of After-Hours Incentives on Services

Effect of After-Hours Services on ED Utilization

Concluding Remarks

Acknowledgements

Footnotes

Supplementary Material

References

Information

Published In

Copyright

History

Keywords:

Mots clés :

Authors

Affiliations

Metrics

Related Content

Citations

View options

PDF

EPUB

Restore your content access

Share

Copy the content Link

Share with email

Share on social media

Request Username

Create a new account

Change Password

Your password must have 8 characters or more and contain 3 of the following:

Password Changed Successfully

Verify Phone

Congrats!