STRATEGIC LEADER ATTRIBUTES, HEALTHCARE SYSTEM CAPACITY, PATIENT-PROVIDER RELATIONAL DYNAMICS, AND PATIENT LOYALTY TO HIV CARE IN AN HIV FACILITY IN ELDORET KENYA
Statisical Analysis Report
By: FC,AK,JW
Background
Patient Loyalty
Patient loyalty is a behavioral impulse that makes a customer engage and remain in the repeated purchase of a particular good or service for a long time (Ngurah et al., 2018). In the present study, patient loyalty to HIV care is a behavioral impulse and commitment of a patient to remain engaged in care regardless of the challenges experienced in care. This involves a commitment to return to the hospital, having confidence in the care, and encouraging others to remain in care.
Strategic Leadership
Strategic leadership involves dealing with problems normally addressed by a firms’ top management team and creating unique relations between management and employees to enhance employees’ performance (Zia-ud-Din et al., 2017). In the present study, adaptive leadership capacity and clinical leader attributes are defined as strategic leader attributes and will be utilized to measure strategic leadership in the context of HIV care where the clinical officer in charge of care oversees the leadership to ensure that the healthcare system environment is favorable for patient engagement and interaction with the provider.
Healthcare System Capacity
According to the World Health Organization (WHO, 2000), healthcare system factors include people, organizations, and actions that have an intention to promote, restore, and maintain health. In the present study, health system factors include factors that aid or constrain healthcare delivery in an HIV care facility. This will be assessed by utilizing the World Health Organization (WHO) health system framework 2000, assessing health system responsiveness to healthcare, and other measures from previous studies.
Relational Dynamics
Relational dynamics is the art of interacting with self and others. It refers to inter-organizational relationships and collaborations that exist within organizational hierarchies and can influence organizational outcomes and decision making (Ozcelik, 2013). In this study, relational dynamics refers to interactions that occur in a healthcare facility involving the patients and providers during HIV care provision, and it is measured by patient trust, relational bonding, and patient-provider communication.
EDA
Patients Sample Size
| Module | n |
|---|---|
| Module 1 | 128 |
| Module 2 | 132 |
| Module 3 | 131 |
## [1] "Total: 391"Providers Sample Size
| Module | n |
|---|---|
| Module 1 | 15 |
| Module 2 | 15 |
| Module 3 | 17 |
## [1] "Total: 47"EFA | Patient
## Parallel analysis suggests that the number of factors = 10 and the number of components = 6##
## Call:
## factanal(x = na.omit(data.fa), factors = Nfacs)
##
## Uniquenesses:
## PL1 PL2 PL3 PL4 PL5 LAC1 LAC2 LAC3 LAC4 LAC5 LAC6 LAC7 HSF1
## 0.25 0.54 0.18 0.50 0.43 0.68 0.61 0.62 0.62 0.58 0.58 0.54 0.66
## HSF2 HSF3 HSF4 HSF5 HSF6 HSF7 HSF8 HSF9 TR1 TR2 TR3 TR4 TR5
## 0.46 0.42 0.35 0.68 0.62 0.74 0.69 0.80 0.57 0.52 0.64 0.50 0.60
## TR6 TR7 TR8 TR9 PPC1 PPC2 PPC3 PPC4 PPC5 PPC6 PPC7 PPC8 PPRB1
## 0.00 0.26 0.25 0.59 0.78 0.72 0.73 0.66 0.66 0.46 0.30 0.55 0.73
## PPRB2 PPRB3 PPRB4 PPRB5 PPRB6 PPRB7 PPRB8 PPRB9 PS1 PS2 PS3 PS4 PS5
## 0.59 0.53 0.45 0.43 0.58 0.21 0.21 0.25 0.50 0.69 0.56 0.58 0.52
## PS6 PS7 PS8 PS9
## 0.40 0.51 0.40 0.57
##
## Loadings:
## Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Factor8 Factor9
## PL1 0.80
## PL2 0.53
## PL3 0.82
## PL4 0.50
## PL5 0.56
## LAC1 0.41
## LAC2 0.39 0.30
## LAC3 0.32 0.38
## LAC4 0.30
## LAC5 0.30 0.46
## LAC6 0.45
## LAC7 0.34 0.44
## HSF1
## HSF2 0.61
## HSF3 0.68
## HSF4 0.74
## HSF5 0.48
## HSF6 0.48
## HSF7 0.43
## HSF8 0.49
## HSF9 0.33
## TR1 0.47
## TR2 0.56
## TR3 0.32
## TR4 0.39 0.34 0.37
## TR5 0.32
## TR6 0.46
## TR7 0.49
## TR8 0.77
## TR9 0.40
## PPC1
## PPC2 0.37
## PPC3
## PPC4 0.38 0.32
## PPC5 0.47
## PPC6 0.69
## PPC7 0.79
## PPC8 0.60
## PPRB1 0.30
## PPRB2 0.44
## PPRB3 0.54
## PPRB4 0.57
## PPRB5 0.68
## PPRB6 0.51
## PPRB7 0.85
## PPRB8 0.86
## PPRB9 0.82
## PS1 0.32 0.60
## PS2 0.42
## PS3 0.50
## PS4 0.58
## PS5 0.63
## PS6 0.71
## PS7 0.64
## PS8 0.71
## PS9 0.52
## Factor10
## PL1
## PL2
## PL3
## PL4
## PL5
## LAC1
## LAC2
## LAC3
## LAC4
## LAC5
## LAC6
## LAC7
## HSF1
## HSF2
## HSF3
## HSF4
## HSF5
## HSF6
## HSF7
## HSF8
## HSF9
## TR1
## TR2
## TR3
## TR4
## TR5 0.32
## TR6 0.82
## TR7 0.59
## TR8
## TR9
## PPC1
## PPC2
## PPC3
## PPC4
## PPC5
## PPC6
## PPC7
## PPC8
## PPRB1
## PPRB2
## PPRB3
## PPRB4
## PPRB5
## PPRB6
## PPRB7
## PPRB8
## PPRB9
## PS1
## PS2
## PS3
## PS4
## PS5
## PS6
## PS7
## PS8
## PS9
##
## Factor1 Factor2 Factor3 Factor4 Factor5 Factor6 Factor7 Factor8
## SS loadings 3.94 3.89 3.60 3.03 2.84 2.78 2.28 1.70
## Proportion Var 0.07 0.07 0.06 0.05 0.05 0.05 0.04 0.03
## Cumulative Var 0.07 0.14 0.20 0.26 0.31 0.36 0.40 0.43
## Factor9 Factor10
## SS loadings 1.55 1.34
## Proportion Var 0.03 0.02
## Cumulative Var 0.46 0.48
##
## Test of the hypothesis that 10 factors are sufficient.
## The chi square statistic is 1722.57 on 1025 degrees of freedom.
## The p-value is 3.01e-38
## Parallel analysis suggests that the number of factors = 2 and the number of components = 1##
## Call:
## factanal(x = na.omit(data.fa), factors = Nfacs, rotation = "promax")
##
## Uniquenesses:
## LAC CLA HSF TR PPC PPRB PS
## 0.26 0.34 0.54 0.29 0.59 0.35 0.57
##
## Loadings:
## Factor1 Factor2
## LAC 0.93
## CLA 0.78
## HSF 0.67
## TR 0.69
## PPC 0.63
## PPRB 0.90
## PS 0.55
##
## Factor1 Factor2
## SS loadings 2.70 1.32
## Proportion Var 0.39 0.19
## Cumulative Var 0.39 0.57
##
## Factor Correlations:
## Factor1 Factor2
## Factor1 1.00 -0.67
## Factor2 -0.67 1.00
##
## Test of the hypothesis that 2 factors are sufficient.
## The chi square statistic is 7.53 on 8 degrees of freedom.
## The p-value is 0.48Correlations | Patient
The positive and statistically significant Pearson Correlation coefficients indicated that all the design variables had direct proportionate magnitude of effects between each other.
Correlations | Providers
As expected, we do see similar behaviour with provider. The positive and statistically significant Pearson Correlation coefficients indicated that all the design variables had direct proportionate magnitude of effects between each other.
Lickert Scale Analysis
Patient
Provider
Bivariate Analysis
Patient
TODO https://www.r-bloggers.com/2016/08/five-ways-to-calculate-internal-consistency/
Socio-economic characteristics | Patients
| Table 1: Socio-economic characteristics | Patients | ||||||
| Patient Loyalty (PL) | ||||||
|---|---|---|---|---|---|---|
|
Total No. 391 |
High No. 265 |
Low No. 126 |
P-value | |||
| Age | 0.004 | |||||
| 18 -29 years | 36 (9.2%) | 23 (63.9%) | 13 (36.1%) | |||
| 30-39 years | 96 (24.6%) | 54 (56.2%) | 42 (43.8%) | |||
| 40-49 years | 133 (34.0%) | 89 (66.9%) | 44 (33.1%) | |||
| > 50 years | 126 (32.2%) | 99 (78.6%) | 27 (21.4%) | |||
| Gender | 0.91 | |||||
| Male | 144 (36.8%) | 97 (67.4%) | 47 (32.6%) | |||
| Female | 247 (63.2%) | 168 (68.0%) | 79 (32.0%) | |||
| Marital Status | 0.47 | |||||
| Married | 230 (58.8%) | 155 (67.4%) | 75 (32.6%) | |||
| Single | 129 (33.0%) | 91 (70.5%) | 38 (29.5%) | |||
| In a relationship | 14 (3.6%) | 7 (50.0%) | 7 (50.0%) | |||
| Divorced/Separated | 18 (4.6%) | 12 (66.7%) | 6 (33.3%) | |||
| Education | 0.002 | |||||
| Primary | 153 (39.1%) | 120 (78.4%) | 33 (21.6%) | |||
| High school | 167 (42.7%) | 104 (62.3%) | 63 (37.7%) | |||
| Vocational | 54 (13.8%) | 33 (61.1%) | 21 (38.9%) | |||
| Graduate | 17 (4.3%) | 8 (47.1%) | 9 (52.9%) | |||
| Income | 0.40 | |||||
| < 100 usd | 252 (64.5%) | 178 (70.6%) | 74 (29.4%) | |||
| 110-300 usd | 78 (19.9%) | 49 (62.8%) | 29 (37.2%) | |||
| 310-500 usd | 15 (3.8%) | 10 (66.7%) | 5 (33.3%) | |||
| > 510 usd | 46 (11.8%) | 28 (60.9%) | 18 (39.1%) | |||
| Household Size | 0.16 | |||||
| 1-5 | 220 (56.3%) | 156 (70.9%) | 64 (29.1%) | |||
| > 6 | 171 (43.7%) | 109 (63.7%) | 62 (36.3%) | |||
| Residential | < 0.0001 | |||||
| Rural | 193 (49.4%) | 136 (70.5%) | 57 (29.5%) | |||
| Urban | 144 (36.8%) | 110 (76.4%) | 34 (23.6%) | |||
| Semi-urban | 54 (13.8%) | 19 (35.2%) | 35 (64.8%) | |||
| ‡ | ||||||
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: ‘Age’, ‘Education’, ‘Residential’ . On the other hand, variables that did not yield statistically significant associations between high and low PL are: ‘Gender’, ‘MStatus’, ‘Income’, ‘Household’ . Variables that yielded marginally significant (p<.1) associations between high and low PL are: .”
Clinical / appointment factors | Patients
| Table 2: Clinical / appointment factors | Patients | ||||||
| Patient Loyalty (PL) | ||||||
|---|---|---|---|---|---|---|
|
Total No. 391 |
High No. 265 |
Low No. 126 |
P-value | |||
| Module (Clinic) | < 0.0001 | |||||
| Module 1 | 128 (32.7%) | 44 (34.4%) | 84 (65.6%) | |||
| Module 2 | 132 (33.8%) | 101 (76.5%) | 31 (23.5%) | |||
| Module 3 | 131 (33.5%) | 120 (91.6%) | 11 (8.4%) | |||
| Traveltime to clinic | 0.66 | |||||
| < 1 hour | 243 (62.1%) | 164 (67.5%) | 79 (32.5%) | |||
| 2-3 hours | 126 (32.2%) | 84 (66.7%) | 42 (33.3%) | |||
| > 4 hours | 22 (5.6%) | 17 (77.3%) | 5 (22.7%) | |||
| Clinic hours | 0.028 | |||||
| 8 hours | 365 (93.4%) | 253 (69.3%) | 112 (30.7%) | |||
| Less than 8 hours | 26 (6.6%) | 12 (46.2%) | 14 (53.8%) | |||
| Visited clinician | < 0.0001 | |||||
| No | 154 (39.4%) | 141 (91.6%) | 13 (8.4%) | |||
| Yes | 237 (60.6%) | 124 (52.3%) | 113 (47.7%) | |||
| Required visit | 0.38 | |||||
| 1-2 | 248 (63.4%) | 165 (66.5%) | 83 (33.5%) | |||
| 3-4 | 115 (29.4%) | 83 (72.2%) | 32 (27.8%) | |||
| > 5 | 28 (7.2%) | 17 (60.7%) | 11 (39.3%) | |||
| Missed visit | 0.0004 | |||||
| 0 (none) | 270 (69.1%) | 199 (73.7%) | 71 (26.3%) | |||
| 1-2 times | 99 (25.3%) | 51 (51.5%) | 48 (48.5%) | |||
| > 3 times | 21 (5.4%) | 14 (66.7%) | 7 (33.3%) | |||
| 5 | 1 (0.3%) | 1 (100.0%) | 0 (0.0%) | |||
| provider seen | 0.34 | |||||
| Clinician | 318 (81.3%) | 211 (66.4%) | 107 (33.6%) | |||
| Nurse | 59 (15.1%) | 42 (71.2%) | 17 (28.8%) | |||
| Pharmacist | 11 (2.8%) | 10 (90.9%) | 1 (9.1%) | |||
| Phlebotomist | 1 (0.3%) | 1 (100.0%) | 0 (0.0%) | |||
| Other (please indicate) | 2 (0.5%) | 1 (50.0%) | 1 (50.0%) | |||
| Provider interaction | 0.42 | |||||
| Clinician | 317 (81.1%) | 213 (67.2%) | 104 (32.8%) | |||
| Nurse | 55 (14.1%) | 36 (65.5%) | 19 (34.5%) | |||
| Pharmacist | 12 (3.1%) | 11 (91.7%) | 1 (8.3%) | |||
| Phlebotomist | 1 (0.3%) | 1 (100.0%) | 0 (0.0%) | |||
| Other (please indicate) | 6 (1.5%) | 4 (66.7%) | 2 (33.3%) | |||
| ‡ | ||||||
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: ‘Module’, ‘Clinichours’, ‘Visitedclinician’, ‘Missedvisit’ . On the other hand, variables that did not yield statistically significant associations between high and low PL are: ‘Traveltime’, ‘Requiredvisit’, ‘providerseen’, ‘Providerinteraction’ . Variables that yielded marginally significant (p<.1) associations between high and low PL are: .”
Study Construct Attributes | Patients
| Table 3: Study Construct Attributes | Patients | ||||||||
| Patient Loyalty (PL) | IC‡ | |||||||
|---|---|---|---|---|---|---|---|---|
|
Total No. 391 |
High No. 265 |
Low No. 126 |
P-value | Cronbach’s α | ||||
| Leader Adaptive Capacity (LAC) | ||||||||
| Mean (SD) | 4.7 (±0.3) | 4.8 (±0.2) | 4.4 (±0.4) | < 0.0001 | 0.79 | |||
| Median (IQR) | 4.7 (4.4 - 5.0) | 4.9 (4.7 - 5.0) | 4.3 (4.1 - 4.6) | |||||
| Clinical Leader Attributes (CLA) | ||||||||
| Mean (SD) | 4.6 (±0.4) | 4.8 (±0.3) | 4.3 (±0.4) | < 0.0001 | 0.79 | |||
| Median (IQR) | 4.7 (4.3 - 4.9) | 4.8 (4.7 - 5.0) | 4.3 (4.0 - 4.6) | |||||
| Health System Capacity (HSF) | ||||||||
| Mean (SD) | 4.6 (±0.4) | 4.7 (±0.3) | 4.4 (±0.4) | < 0.0001 | 0.81 | |||
| Median (IQR) | 4.8 (4.4 - 4.9) | 4.8 (4.7 - 5.0) | 4.4 (4.2 - 4.7) | |||||
| Trust (TR) in clinician and system | ||||||||
| Mean (SD) | 4.3 (±0.6) | 4.6 (±0.4) | 3.8 (±0.4) | < 0.0001 | 0.76 | |||
| Median (IQR) | 4.6 (3.8 - 4.9) | 4.8 (4.3 - 4.9) | 3.7 (3.6 - 4.0) | |||||
| Patient Provider Communication (PPC) | ||||||||
| Mean (SD) | 4.7 (±0.4) | 4.8 (±0.3) | 4.5 (±0.4) | < 0.0001 | 0.79 | |||
| Median (IQR) | 4.8 (4.5 - 5.0) | 4.9 (4.6 - 5.0) | 4.5 (4.2 - 4.8) | |||||
| Patient Provider Bonding (PPRB) | ||||||||
| Mean (SD) | 4.0 (±0.7) | 4.3 (±0.6) | 3.6 (±0.6) | < 0.0001 | 0.79 | |||
| Median (IQR) | 4.0 (3.5 - 4.7) | 4.4 (3.7 - 4.8) | 3.6 (3.1 - 4.0) | |||||
| Patient service (PS) | ||||||||
| Mean (SD) | 4.7 (±0.4) | 4.8 (±0.2) | 4.5 (±0.4) | < 0.0001 | 0.84 | |||
| Median (IQR) | 4.9 (4.6 - 5.0) | 5.0 (4.8 - 5.0) | 4.4 (4.2 - 4.9) | |||||
| Patient Loyalty (PL) | ||||||||
| Mean (SD) | 4.7 (±0.4) | 5.0 (±0.1) | 4.2 (±0.4) | < 0.0001 | 0.87 | |||
| Median (IQR) | 5.0 (4.6 - 5.0) | 5.0 (5.0 - 5.0) | 4.2 (4.0 - 4.4) | |||||
| IC‡ : Internal Consistency | ||||||||
IC / Internal Consistency, measures whether several items that propose to measure the same general construct produce similar scores. Internal consistency is usually measured with Cronbach’s alpha, a statistic calculated from the pairwise correlations between items. Here is a range-map that can be used to interprete
- 0.9 ≤ α Excellent
- 0.8 ≤ α < 0.9 Good
- 0.7 ≤ α < 0.8 Acceptable
- 0.6 ≤ α < 0.7 Questionable
- 0.5 ≤ α < 0.6 Poor
- α < 0.5 Unacceptable
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: ‘LAC’, ‘CLA’, ‘HSF’, ‘TR’, ‘PPC’, ‘PPRB’, ‘PS’, ‘PL’ . On the other hand, variables that did not yield statistically significant associations between high and low PL are: . Variables that yielded marginally significant (p<.1) associations between high and low PL are: .”
Provider
Socio-economic characteristics | Providers
| Table 4: Socio-economic characteristics | Providers | ||||||
| Patient Loyalty (PL) | ||||||
|---|---|---|---|---|---|---|
|
Total No. 47 |
High No. 30 |
Low No. 17 |
P-value | |||
| Age | 0.19 | |||||
| 18 -30 years | 2 (4.3%) | 0 (0.0%) | 2 (100.0%) | |||
| 31-40 years | 17 (36.2%) | 10 (58.8%) | 7 (41.2%) | |||
| 41-50 years | 26 (55.3%) | 18 (69.2%) | 8 (30.8%) | |||
| Adult | 2 (4.3%) | 2 (100.0%) | 0 (0.0%) | |||
| Gender | 0.066 | |||||
| Male | 26 (55.3%) | 20 (76.9%) | 6 (23.1%) | |||
| Female | 21 (44.7%) | 10 (47.6%) | 11 (52.4%) | |||
| Education | 0.21 | |||||
| College | 30 (63.8%) | 21 (70.0%) | 9 (30.0%) | |||
| Undergraduate | 12 (25.5%) | 5 (41.7%) | 7 (58.3%) | |||
| Post-graduate | 5 (10.6%) | 4 (80.0%) | 1 (20.0%) | |||
| Experience | 0.40 | |||||
| < 3 years | 6 (12.8%) | 3 (50.0%) | 3 (50.0%) | |||
| 4-7 years | 16 (34.0%) | 9 (56.2%) | 7 (43.8%) | |||
| > 7 years | 25 (53.2%) | 18 (72.0%) | 7 (28.0%) | |||
| Income | 0.79 | |||||
| <50000 | 17 (36.2%) | 12 (70.6%) | 5 (29.4%) | |||
| 51000-100000 | 9 (19.1%) | 6 (66.7%) | 3 (33.3%) | |||
| 101000-150000 | 11 (23.4%) | 7 (63.6%) | 4 (36.4%) | |||
| > 151000 | 10 (21.3%) | 5 (50.0%) | 5 (50.0%) | |||
| ‡ | ||||||
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: . On the other hand, variables that did not yield statistically significant associations between high and low PL are: ‘Age’, ‘Education’, ‘Experience’, ‘Income’ . Variables that yielded marginally significant (p<.1) associations between high and low PL are: ‘Gender’ .”
Provider training / professional / profficiency attributes
| Table 5: Provider training / profficiency attributes | ||||||
| Patient Loyalty (PL) | ||||||
|---|---|---|---|---|---|---|
|
Total No. 47 |
High No. 30 |
Low No. 17 |
P-value | |||
| Module (Clinic) | 0.31 | |||||
| Module 1 | 15 (31.9%) | 12 (80.0%) | 3 (20.0%) | |||
| Module 2 | 15 (31.9%) | 9 (60.0%) | 6 (40.0%) | |||
| Module 3 | 17 (36.2%) | 9 (52.9%) | 8 (47.1%) | |||
| Formal Education Leadership | 0.11 | |||||
| Yes | 32 (68.1%) | 23 (71.9%) | 9 (28.1%) | |||
| No | 15 (31.9%) | 7 (46.7%) | 8 (53.3%) | |||
| Formal Education Clinicaleadership | 0.76 | |||||
| Yes | 25 (53.2%) | 15 (60.0%) | 10 (40.0%) | |||
| No | 22 (46.8%) | 15 (68.2%) | 7 (31.8%) | |||
| Profession | 0.19 | |||||
| Clinician | 16 (34.0%) | 8 (50.0%) | 8 (50.0%) | |||
| Nurse | 12 (25.5%) | 7 (58.3%) | 5 (41.7%) | |||
| Other (indicate | 19 (40.4%) | 15 (78.9%) | 4 (21.1%) | |||
| Working Unit | 0.51 | |||||
| Clnical | 20 (42.6%) | 11 (55.0%) | 9 (45.0%) | |||
| Nursing | 14 (29.8%) | 9 (64.3%) | 5 (35.7%) | |||
| Other (indicate) | 13 (27.7%) | 10 (76.9%) | 3 (23.1%) | |||
| Received Performance Incetives | 0.011 | |||||
| Yes | 29 (61.7%) | 23 (79.3%) | 6 (20.7%) | |||
| No | 18 (38.3%) | 7 (38.9%) | 11 (61.1%) | |||
| Recognition | 0.73 | |||||
| Yes | 35 (74.5%) | 23 (65.7%) | 12 (34.3%) | |||
| No | 12 (25.5%) | 7 (58.3%) | 5 (41.7%) | |||
| Delivery of Care | 0.36 | |||||
| Yes | 46 (97.9%) | 30 (65.2%) | 16 (34.8%) | |||
| No | 1 (2.1%) | 0 (0.0%) | 1 (100.0%) | |||
| Professional Development | 1.0 | |||||
| Yes | 43 (91.5%) | 27 (62.8%) | 16 (37.2%) | |||
| No | 4 (8.5%) | 3 (75.0%) | 1 (25.0%) | |||
| Role Sharing | 0.60 | |||||
| Yes | 45 (95.7%) | 29 (64.4%) | 16 (35.6%) | |||
| No | 1 (2.1%) | 0 (0.0%) | 1 (100.0%) | |||
| 5 | 1 (2.1%) | 1 (100.0%) | 0 (0.0%) | |||
| ‡ | ||||||
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: ‘ReceivedPerformanceIncetives’ . On the other hand, variables that did not yield statistically significant associations between high and low PL are: ‘Module’, ‘FormalEducationLeadership’, ‘FormalEducationClinicaleadership’, ‘Profession’, ‘WorkingUnit’, ‘Recognition’, ‘DeliveryofCare’, ‘ProfessionalDevelopment’, ‘RoleSharing’ . Variables that yielded marginally significant (p<.1) associations between high and low PL are: .”
Study Constructs | Providers
| Table 6: Study Construct Attributes | Providers | ||||||||
| Patient Loyalty (PL) | IC‡ | |||||||
|---|---|---|---|---|---|---|---|---|
|
Total No. 47 |
High No. 30 |
Low No. 17 |
P-value | Cronbach’s α | ||||
| Leader Adaptive Capacity (LAC) | ||||||||
| Mean (SD) | 4.2 (±0.8) | 4.4 (±0.8) | 3.8 (±0.6) | 0.0005 | 0.89 | |||
| Median (IQR) | 4.3 (3.9 - 4.8) | 4.7 (4.2 - 5.0) | 4.0 (3.7 - 4.1) | |||||
| Clinical Leader Attributes (CLA) | ||||||||
| Mean (SD) | 4.4 (±0.6) | 4.6 (±0.5) | 4.0 (±0.6) | 0.0003 | 0.89 | |||
| Median (IQR) | 4.6 (4.2 - 4.9) | 4.8 (4.4 - 5.0) | 4.3 (3.7 - 4.4) | |||||
| Health System Capacity (HSF) | ||||||||
| Mean (SD) | 4.5 (±0.5) | 4.7 (±0.3) | 4.2 (±0.6) | 0.004 | 0.86 | |||
| Median (IQR) | 4.7 (4.3 - 4.9) | 4.8 (4.6 - 5.0) | 4.2 (4.0 - 4.7) | |||||
| Trust (TR) in clinician and system | ||||||||
| Mean (SD) | 4.2 (±0.5) | 4.4 (±0.3) | 3.9 (±0.7) | 0.001 | 0.75 | |||
| Median (IQR) | 4.3 (4.0 - 4.6) | 4.4 (4.2 - 4.6) | 4.0 (3.4 - 4.2) | |||||
| Patient Provider Communication (PPC) | ||||||||
| Mean (SD) | 4.6 (±0.6) | 4.8 (±0.3) | 4.2 (±0.7) | 0.001 | 0.92 | |||
| Median (IQR) | 4.8 (4.2 - 5.0) | 4.9 (4.6 - 5.0) | 4.4 (3.6 - 4.8) | |||||
| Patient Provider Bonding (PPRB) | ||||||||
| Mean (SD) | 3.9 (±0.6) | 4.1 (±0.4) | 3.6 (±0.8) | 0.070 | 0.77 | |||
| Median (IQR) | 4.0 (3.7 - 4.4) | 4.1 (3.8 - 4.6) | 3.7 (2.9 - 4.2) | |||||
| Patient service (PS) | ||||||||
| Mean (SD) | 4.5 (±0.6) | 4.7 (±0.3) | 4.0 (±0.7) | 0.0002 | 0.93 | |||
| Median (IQR) | 4.7 (4.1 - 4.9) | 4.9 (4.5 - 5.0) | 4.1 (3.8 - 4.6) | |||||
| Patient Loyalty (PL) | ||||||||
| Mean (SD) | 4.5 (±0.6) | 4.9 (±0.2) | 3.8 (±0.6) | < 0.0001 | 0.87 | |||
| Median (IQR) | 4.6 (4.0 - 5.0) | 5.0 (4.6 - 5.0) | 4.0 (3.6 - 4.0) | |||||
| IC‡ : Internal Consistency | ||||||||
IC / Internal Consistency, measures whether several items that propose to measure the same general construct produce similar scores. Internal consistency is usually measured with Cronbach’s alpha, a statistic calculated from the pairwise correlations between items. Here is a range-map that can be used to interprete
- 0.9 ≤ α Excellent
- 0.8 ≤ α < 0.9 Good
- 0.7 ≤ α < 0.8 Acceptable
- 0.6 ≤ α < 0.7 Questionable
- 0.5 ≤ α < 0.6 Poor
- α < 0.5 Unacceptable
[1] “Some of the variables that yielded statistically significant (p<.05) associations between high and low PL are: ‘LAC’, ‘CLA’, ‘HSF’, ‘TR’, ‘PPC’, ‘PS’, ‘PL’ . On the other hand, variables that did not yield statistically significant associations between high and low PL are: . Variables that yielded marginally significant (p<.1) associations between high and low PL are: ‘PPRB’ .”
Multivariate Analysis
Patient
Model Validation Checks | Patient
Before model interpretation, let’s check on model accuracies.
##
## Call:
## roc.formula(formula = PL_Status ~ final.logit$fitted.values, data = model.df, plot = TRUE, grid = TRUE, print.auc = TRUE, show.thres = TRUE, ci = TRUE, boot.n = 100, ci.alpha = 0.9, stratified = FALSE, main = "ROC Curve", col = "blue")
##
## Data: final.logit$fitted.values in 126 controls (PL_Status 0) < 265 cases (PL_Status 1).
## Area under the curve: 0.9511
## 95% CI: 0.928-0.9742 (DeLong)An AUC of > 80% generally indicates a good model fit. In this case, we have an AUC of over 95%, with very precise CI
Model Tabulation
Patient Loyalty Logit Model | Patient | ||||||
| Unadjusted Estimates | Adjusted Estimates | ||||
Characteristic | OR1 | 95% CI1 | p-value | OR1 | 95% CI1 | p-value |
Age | ||||||
18 -29 years | — | — | ||||
30-39 years | 0.73 | (0.32, 1.59) | 0.4 | 1.18 | (0.28, 4.78) | 0.8 |
40-49 years | 1.14 | (0.52, 2.44) | 0.7 | 1.98 | (0.51, 7.57) | 0.3 |
> 50 years | 2.07 | (0.91, 4.60) | 0.075 | 1.72 | (0.44, 6.48) | 0.4 |
Gender | ||||||
Male | — | — | ||||
Female | 1.03 | (0.66, 1.60) | 0.9 | 0.68 | (0.29, 1.55) | 0.4 |
Education | ||||||
Primary | — | — | ||||
High school | 0.45 | (0.27, 0.74) | 0.002 | 0.38 | (0.15, 0.93) | 0.037 |
Vocational | 0.43 | (0.22, 0.85) | 0.014 | 1.61 | (0.48, 5.46) | 0.4 |
Graduate | 0.24 | (0.09, 0.69) | 0.007 | 0.20 | (0.03, 1.15) | 0.071 |
Module | ||||||
Module 1 | — | — | ||||
Module 2 | 6.22 | (3.65, 10.8) | <0.001 | 3.01 | (1.23, 7.63) | 0.017 |
Module 3 | 20.8 | (10.5, 44.7) | <0.001 | 4.34 | (1.14, 16.7) | 0.031 |
Residential | ||||||
Rural | — | — | ||||
Urban | 1.36 | (0.83, 2.24) | 0.2 | 0.59 | (0.23, 1.51) | 0.3 |
Semi-urban | 0.23 | (0.12, 0.43) | <0.001 | 0.25 | (0.08, 0.77) | 0.017 |
Traveltime | ||||||
< 1 hour | — | — | ||||
2-3 hours | 0.96 | (0.61, 1.53) | 0.9 | 0.64 | (0.25, 1.62) | 0.3 |
> 4 hours | 1.64 | (0.62, 5.13) | 0.3 | 1.09 | (0.20, 6.78) | >0.9 |
Visitedclinician | ||||||
No | — | — | ||||
Yes | 0.10 | (0.05, 0.18) | <0.001 | 0.33 | (0.11, 0.91) | 0.036 |
Clinichours | ||||||
8 hours | — | — | ||||
Less than 8 hours | 0.38 | (0.17, 0.85) | 0.018 | 0.23 | (0.06, 0.84) | 0.030 |
LAC | ||||||
High | — | — | ||||
Low | 0.08 | (0.05, 0.13) | <0.001 | 0.12 | (0.05, 0.28) | <0.001 |
CLA | ||||||
High | — | — | ||||
Low | 0.09 | (0.05, 0.14) | <0.001 | 0.40 | (0.17, 0.92) | 0.032 |
HSF | ||||||
High | — | — | ||||
Low | 0.11 | (0.07, 0.18) | <0.001 | 0.49 | (0.21, 1.14) | 0.10 |
TR | ||||||
High | — | — | ||||
Low | 0.04 | (0.02, 0.08) | <0.001 | 0.09 | (0.03, 0.26) | <0.001 |
PPC | ||||||
High | — | — | ||||
Low | 0.20 | (0.12, 0.31) | <0.001 | 1.26 | (0.54, 3.09) | 0.6 |
PPRB | ||||||
High | — | — | ||||
Low | 0.16 | (0.10, 0.26) | <0.001 | 2.22 | (0.84, 6.18) | 0.12 |
PS | ||||||
High | — | — | ||||
Low | 0.14 | (0.09, 0.23) | <0.001 | 0.37 | (0.15, 0.87) | 0.023 |
1OR = Odds Ratio, CI = Confidence Interval | ||||||
In general, all the design variables were statistically significant in unadjusted regression. After adjustment for clinical, social-demographic, and other design covariates we see PPC, and PPRB becoming insignificant; TR becoming marginally insignificant, while the rest of the design remaining statistically significant.
| variable_name | interpretation | Significant |
|---|---|---|
| Age30-39 years | The odds of Patient Loyalty in age30-39 years is 1.18 ( 0.28 , 4.78 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| Age40-49 years | The odds of Patient Loyalty in age40-49 years is 1.98 ( 0.51 , 7.57 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| Age> 50 years | The odds of Patient Loyalty in age> 50 years is 1.72 ( 0.44 , 6.48 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| GenderFemale | The odds of Patient Loyalty in genderfemale is 0.68 ( 0.29 , 1.55 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| EducationHigh school | The odds of Patient Loyalty in educationhigh school is 0.38 ( 0.15 , 0.93 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| EducationVocational | The odds of Patient Loyalty in educationvocational is 1.61 ( 0.48 , 5.46 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| EducationGraduate | The odds of Patient Loyalty in educationgraduate is 0.2 ( 0.03 , 1.15 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| ModuleModule 2 | The odds of Patient Loyalty in modulemodule 2 is 3.01 ( 1.23 , 7.63 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| ModuleModule 3 | The odds of Patient Loyalty in modulemodule 3 is 4.34 ( 1.14 , 16.67 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| ResidentialUrban | The odds of Patient Loyalty in residentialurban is 0.59 ( 0.23 , 1.51 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| ResidentialSemi-urban | The odds of Patient Loyalty in residentialsemi-urban is 0.25 ( 0.08 , 0.77 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| Traveltime2-3 hours | The odds of Patient Loyalty in traveltime2-3 hours is 0.64 ( 0.25 , 1.62 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| Traveltime> 4 hours | The odds of Patient Loyalty in traveltime> 4 hours is 1.09 ( 0.2 , 6.78 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| VisitedclinicianYes | The odds of Patient Loyalty in visitedclinicianyes is 0.33 ( 0.11 , 0.91 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| ClinichoursLess than 8 hours | The odds of Patient Loyalty in clinichoursless than 8 hours is 0.23 ( 0.06 , 0.84 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| LACLow | The odds of Patient Loyalty in laclow is 0.12 ( 0.05 , 0.28 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| CLALow | The odds of Patient Loyalty in clalow is 0.4 ( 0.17 , 0.92 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| HSFLow | The odds of Patient Loyalty in hsflow is 0.49 ( 0.21 , 1.14 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| TRLow | The odds of Patient Loyalty in trlow is 0.09 ( 0.03 , 0.26 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
| PPCLow | The odds of Patient Loyalty in ppclow is 1.26 ( 0.54 , 3.09 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| PPRBLow | The odds of Patient Loyalty in pprblow is 2.22 ( 0.84 , 6.18 ) times higher than the odds of Patient Loyalty in the reference level, fixing all else constant. | No |
| PSLow | The odds of Patient Loyalty in pslow is 0.37 ( 0.15 , 0.87 ) times lower than the odds of Patient Loyalty in the reference level, fixing all else constant. | Yes |
Model visualization | Patient
The precision plot of the findings indicates the range within which the predicted estimates (OR) of population may lie.
A confidence interval help us evaluate the range of estimated values within which the actual “ground-truth” result is found. Essentially, a CI of 95% means that if a trial was repeated an infinite number of times, 95% of the results would fall within this range of values. In our analysis we see very precise estimates, i.e., narrow confidence intervals.
The average marginal effect gives you an effect on the probability scale, i.e. a number between 0 and 1. It is the average change in probability when each predictor increases by one unit, holding the other variables constant/or at reference level for categorical variables. Since a probit is a non-linear model, that effect will differ from individual to individual. What the average marginal effect does is compute it for each individual and than compute the average. To get the effect on the percentage you need to multiply by a 100, so the chances of patient becoming loyal, increases by by x percentage points.
Multicollinearity diagnostics VIF
| GVIF | Df | GVIF^(1/(2*Df)) | |
|---|---|---|---|
| Age | 2.027296 | 3 | 1.125001 |
| Gender | 1.300586 | 1 | 1.140432 |
| Education | 1.757312 | 3 | 1.098520 |
| Module | 2.383592 | 2 | 1.242533 |
| Residential | 1.894113 | 2 | 1.173144 |
| Traveltime | 1.829782 | 2 | 1.163054 |
| Visitedclinician | 1.445274 | 1 | 1.202195 |
| Clinichours | 1.151880 | 1 | 1.073257 |
| LAC | 1.598828 | 1 | 1.264447 |
| CLA | 1.404072 | 1 | 1.184935 |
| HSF | 1.367033 | 1 | 1.169202 |
| TR | 1.800071 | 1 | 1.341667 |
| PPC | 1.482779 | 1 | 1.217694 |
| PPRB | 1.806564 | 1 | 1.344085 |
| PS | 1.426672 | 1 | 1.194434 |
Multicollinearity is present if GVIF >10. In our model GVIFs are approximately ~1 suggesting we do not have Multicollinearity.
Provider
TODO (In progress)