The seven health centers (with 23 study clinics) had patient populations ranging in size from 3,584 patients to 22,286 patients during the implementation period (Table 2). The rate of uninsured visits varied across clinics, from 13.9% to 52.8%. In all but one health center, more than 40% of encounters were Medicaid-insured. Characteristics of the patient population that health centers served (e.g., race and ethnicity) varied.
Table 2. Characteristics of participating health centers
|
ARM 1
(basic education material)
|
ARM 2
(basic education materials + facilitation)
|
|
|
Health Center A
|
Health Center B
|
Health Center C
|
Health Center D
|
Health Center E
|
Health Center F
|
Health Center G
|
Number active patientsa
|
3,584
|
20,830
|
4,368
|
4,334
|
13,301
|
4,020
|
22,286
|
Number clinics
|
3
|
4
|
3
|
1
|
5
|
1
|
6
|
% Uninsured visits
|
17.5
|
20.1
|
20.2
|
52.8
|
13.9
|
19.8
|
17.8
|
% Medicaid insured visits
|
42.7
|
52.3
|
47.1
|
41.5
|
61.7
|
26.4
|
51.8
|
Urbanicityb
|
Urban
|
Urban
|
Mixed
|
Urban
|
Urban
|
Rural
|
Urban
|
% Nonwhite
|
4.8
|
24.1
|
2.4
|
11.0
|
21.0
|
1.0
|
2.8
|
% Hispanic
|
6.5
|
26.3
|
26.9
|
49.3
|
61.4
|
2.5
|
18.1
|
Median age (years)
|
52.6
|
33.2
|
37.7
|
43.5
|
46.8
|
53.1
|
46.6
|
% Income <138% FPLc
|
70.1
|
78.4
|
39.1
|
85.0
|
53.2
|
31.6
|
37.5
|
|
|
|
|
|
|
|
|
|
a) Active patients defined as individuals with a ambulatory visit during the study period (September, 2016-March, 2018)
b) Urbanicity defined as all clinic sites located in urban areas (≥2500 residents), all clinics located in rural areas (<2500 residents), or mixed (clinics located in both urban and rural areas). Urban and rural areas determined according to the 2010 US Census.
c) FPL: Federal Poverty Level
Most clinics utilized the tool
From Table 3, five of seven health centers recorded tool use. One health center in Arm 1 (D) and one health center in Arm 2 (E) did not record any tool use. Health Center G (Arm 2) had the highest rate of tool utilization, with over five times more unique patients with tool use and just over 4 times more instances of tool use compared to the next closest health center. The population for whom the tool was used was generally similar across health centers, and most were uninsured or Medicaid-insured patients. However, at Health Center A, 44% of individuals who received insurance support with the enrollment tool never became patients at that health center during the study period, demonstrating that their health center utilized the tool to engage in more community enrollment support than other CHCs. Though most tool utilization assisted one person per encounter, many unique instances of tool use assisted more than one person. At health center C, 49% of tool use instances assisted multiple individuals while at health center G, only 22% of tool use instances assisted multiple people.
Table 3. Tool use by health center during the implementation period, September, 2016-March, 2018
|
ARM 1
|
ARM 2
|
|
Health Center A
|
Health Center B
|
Health Center C
|
Health Center D
|
Health Center E
|
Health Center F
|
Health Center G
|
Unique patients with tool use
|
662
|
279
|
1,600
|
0
|
0
|
432
|
8,403
|
Total instancesa of tool use
|
747
|
374
|
3,047
|
0
|
0
|
609
|
13,068
|
Patient Status
Established patient
New patient
Never patient
|
329 (49.7)
41 (6.2)
292 (44.1)
|
220 (78.9)
28 (10.0)
31 (11.1)
|
1,418 (88.6)
50 (3.1)
132 (8.2)
|
0 (0.0)
0 (0.0)
0 (0.0)
|
0 (0.0)
0 (0.0)
0 (0.0)
|
402 (93.1)
11 (2.6)
19 (4.4)
|
7,382 (87.9)
243 (2.9)
778 (9.3)
|
Total # persons assisted
No Information
1
>1
|
1 (0.2)
452 (68.3)
209 (31.6)
|
2 (0.7)
319 (78.5)
58 (20.8)
|
66 (4.1)
750 (46.9)
784 (49.0)
|
0 (0.0)
0 (0.0)
0 (0.0)
|
0 (0.0)
0 (0.0)
0 (0.0)
|
12 (2.8)
275 (63.7)
145 (33.6)
|
146 (1.7)
6,399 (76.2)
1,858 (22.1)
|
Insurance type prior to first tool use
Medicaid
Medicare
Private
Other publicb
Uninsured
No prior visits/Missing
|
183 (27.6)
15 (2.3)
28 (4.2)
16 (2.4)
64 (9.7)
356 (53.8)
|
46 (16.5)
4 (1.4)
19 (6.8)
1 (0.4)
114 (40.9)
95 (34.1)
|
779 (48.7)
31 (1.9)
240 (15.0)
40 (2.5)
315 (19.7)
195 (12.2)
|
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
|
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
|
296 (68.5)
7 (1.6)
51 (11.8)
0 (0.0)
37 (8.6)
41 (9.5)
|
5,543 (66.0)
85 (1.0)
352 (4.2)
72 (0.9)
1,171(14.0)
1180 (14.0)
|
Number of staff using the tool
|
3
|
5
|
8
|
0
|
0
|
5
|
10
|
a) Instances of tool use defined as unique patient-dates of tool use
b) Other Public Insurance includes publically-funded coverage sources typically covering limited services (e.g., breast and cervical cancer early detection program; title X contraceptive care) or available to specific populations (e.g., VA and Tricare, Indian Health Service, grant programs for migrant/seasonal workers, and care for the homeless or individuals living with HIV/AIDS)
For patients at high risk of uninsurance, Arm 2 health centers used the tool more frequently than Arm 1 health centers
Among the population we defined as at highest risk of insurance discontinuity – those with at least one Medicaid-insured or uninsured ambulatory visit during the study period (n=51,656) – a total of 6,602 (12.8%) unique patients received insurance support with the enrollment tool (Table 4). Overall, the rate of patients with at least one Medicaid-insured or uninsured visit who received support with the enrollment tool was significantly higher in Arm 2 health centers compared to those in Arm 1 (20.0% vs 4.7%, p <0.01, RR= 4.27 95%, CI = 4.01-4.56). Table 4 also demonstrates variability in rate of tool use between health centers with a higher rate of tool use by Health Center G (33%) and a much lower rate of tool use at health center B (1.5%).
Table 4. Comparison of tool utilization (by health center and study arm) among patients with ≥1 Medicaid-covered or uninsured ambulatory visit during the implementation period (September,2016-March, 2018)
|
Arm 1
|
Arm 2
|
RRa (95% CI)
|
|
Health Center A
|
Health Center B
|
Health Center C
|
Health Center D
|
Health Center E
|
Health Center F
|
Health Center G
|
|
Number patients with tool use
|
211
|
226
|
696
|
0
|
0
|
347
|
5,122
|
|
Total number of patients
|
2,155
|
15,066
|
2,942
|
4,087
|
10,049
|
1,856
|
15,501
|
|
Percent of patients with tool use (by CHC)
|
9.8%
|
1.5%
|
23.7%
|
0.0%
|
0.0%
|
18.7%
|
33.0%
|
|
Percent of patients with tool use (by arm)
|
4.7%
|
20%
|
4.27
(4.01, 4.56)
p<0.001
|
a) RR = Rate Ratio comparing study arms; CI = Confidence Interval
Tool utilization varied over time, with similar tool utilization rates in study Arms 1 and 2 by the end of the study
Monthly tool utilization rates varied over time for health centers in both study arms. Arm 2 CHCs had a high rate of tool use initially which peaked sharply in March of 2017, and subsequently decreased with multiple smaller surges, ending at a rate similar to Arm 1. Utilization in Arm 1 increased sharply in month 3 of the project and remained approximately steady afterward. (Figure 1)
Stratifying the total instances of tool utilization by health center (Figure 2) demonstrates that the shape of the Arm 2 line in Figure 1 is similar to the shape of the line of tool utilization for Health Center G. In some health centers, small peaks of tool utilization were evident during open enrollment periods, most notably at Health Center C. Variability of tool use over time was impacted by multiple factors including leadership engagement and implementation climate (described below).
Perceived Relative Advantage played an important role in implementation
Three CFIR elements were identified as influencing implementation of the enrollment tool at the health centers: relative advantage, implementation climate, and leadership engagement. Table 5 provides a definition of these three CFIR elements and corresponding qualitative data examples to illustrate how health centers varied from low to high on each element. Table 6 shows the variation of the CFIR elements across health centers.
Health centers reported that the enrollment tool’s relative advantage over existing systems was one main factor driving tool utilization. The organizations with the most favorable views of the tool either had no previous system in place for tracking enrollment work, or systems they considered disorganized. The biggest advantages of the enrollment tool were its utility in tracking enrollment applications during the period between an application’s submission and acceptance, and its usefulness for federal HRSA reporting. For example, Health Center E did not use the enrollment tool and instead used a different insurance tracking tool outside of their EHR system that better suited the needs of their staff (who did not always have access to the EHR). Health Center B started using the tool more after discovering its advantage of fulfilling HRSA reporting. Health Center G found the tool most beneficial when paired with Medicaid date of coverage data (the specific date on which a patient’s Medicaid insurance would expire) which they were intermittently able to acquire from a state-level partner and use for proactive patient outreach.
Some health centers reported that implementation climate also influenced tool utilization. Health centers with high implementation support, usually from technical staff or leadership, found that this support promoted consistency in tool use across assister teams; these health centers commonly used audit and feedback mechanisms to address issues and inconsistencies in enrollment tool use as they arose. This support also helped assister teams initially understand where tool use might best fit within their workflow, and how it might benefit their work. For example, a collaborative climate of partnership with the local Accountable Care Organization to receive health insurance coverage dates among patients in Health Center G occurred at the same time as the dramatic increase in tool utilization at Health Center G in early 2017. When this partnership was not sustained, there was a subsequent decrease in tool utilization. Health centers with a less favorable implementation climate, such as Health Center C, had assister staff members who found the tool to be burdensome, and some reported to be doubling up their work.
Finally, leadership engagement impacted how the tool was initially received at a health center and could drive tool use. Engaged upper and middle management leaders were able to facilitate buy-in for the tool from assister teams, especially when leaders conveyed excitement regarding how the tool might benefit their own work and/or conveyed the expectation to assisters that they use the tool, and that their work is part of a larger goal. This was exemplified in Health Center C which had a competing tool developed by one of the assisters, but the directive from leadership fostered use the enrollment tool. Conversely, unengaged health center leaders had little knowledge of the work that assisters do, nor knowledge of the enroll tool. In the case of Health Center D, leadership turned over and left no one at the health center engaged in tool implementation. As such, this health center did not use the tool at all.
Table 5. CFIR Elements and implementation: qualitative examples
|
Relative Advantage
Stakeholders’ perception of the advantage of implementing the intervention versus an alternative solution
|
High
|
“You know, [the enrollment tool] sure beats the notes that you'd have to put in. I mean before it was, you know, note after note. Now there's a place for a comment and there's a place for what you did, and you just click different things. It's really quick… I think you tend to capture more of the people you helped.” Enrollment assister Interview, Health Center A
|
Low
|
“We talk to the assister about what she would like to see in the [enrollment] tool. She wants to have multiple boxes so that each family member, their DOB, and Medicaid number can be all the same form, and she would like a tool that would be good for tracking. Her supervisor asks what she likes better, the [enrollment] tool or the Access database system they used previously. The assister immediately and emphatically says that the Access database was better.” Fieldnotes, Health Center C
|
Implementation Climate
The absorptive capacity for change, shared receptivity of involved individuals to an intervention, and the extent to which use of that intervention will be rewarded, supported, and expected within their organization
|
Strong
|
“…Our EPIC clinic applications team really owned the training of the [enrollment] tool. So we sat down in a group [with assisters], and we had a guide…a step by step, here’s what you do. And then we logged into computers, all in the same room, and practiced with it as well…” Operations Manager Interview, Health Center G
|
Weak
|
“I don't know if we got an email or what it was. The [EHR specialist] said that starting October 1st…we would have to use it so it made it sound like it was not an option, and I will be honest, we were not happy about making it, but we made the changes and so we did start using it as of August 1st. We did have a lot of hiccups in the beginning… I didn't read the instructions as thoroughly as I should have, but it wasn't well received in the beginning.” Enrollment assister Interview, Health Center C
|
Leadership engagement
Commitment, involvement, and accountability of leaders and managers with the implementation.
|
Strong
|
“Our goal, or hope as an FQHC is to provide care for every single Medicaid-covered person in the county... With the alternative payment model and with some of the, sort of incentives or quality metrics that [our Accountable Care Organization] has put in front of us, we definitely need to be doing more outreach. And I think that's where the [enrollment tool] helps a lot…Yeah, it’s very helpful to be able to track that and-, and keep ahead of that because, um, the Medicaid system is our best payer.” Chief Financial Officer, Interview, Health Center F
|
Weak
|
“I [sighs] am very upfront and open about the fact that the outreach worker position is an area that I don’t know much about. It was put under me kind of as an afterthought. …Someday I would like to know more about all of that stuff, and what the tools look like and what the process is and where we can go from there. But right now, it’s just like – it’s the next thing on my agenda. Behavioral Health Director [and head of department that includes enrollment assisters] Interview, Health Center A
|
Table 6. CFIR element ratings by health center
|
|
Arm 1
|
Arm 2
|
|
|
Health Center A
|
Health Center B
|
Health Center C
|
Health Center D
|
Health Center E
|
Health Center F
|
Health Center G
|
|
Relative Advantage
|
High
|
Low / High**
|
Low
|
No data
|
Low
|
High
|
High
|
|
Implementation Climate
|
High
|
High
|
Low
|
No data
|
High
|
Low
|
High
|
|
Leadership Engagement
|
Low
|
High
|
High
|
None*
|
Low
|
High
|
High
|
|
* Leaders that agreed to implement the tool left the organization; new leaders were unengaged.
** This practice did not see the advantage of this tool until team members discovered its HRSA reporting functionality. These additions changed practice members perceptions of relative advantage of using this tool from low to high.