Factors associated with defensive medicine practice in United Arab Emirates: A cross-sectional study with multivariate analysis

Abstract

Health services institutes worldwide are trying to reduce defensive medical practice to limit its negative impact on patient care. We evaluated the factors associated with this defensive medical practice among medical professionals in the United Arab Emirates. This study deployed multivariate logistic regression analysis. Defensive medical practice was defined according to the responses given to questions about potentially unnecessary referral, testing, and additional care in a cross-sectional 23-item questionnaire administered to medical professionals after obtaining ethical committee approval. The factors evaluated were: age, gender, medical specialty, job grade, years of practice, country of medical graduation, country of specialty board, current practice in hospital or private sector, feeling supported by workplace staff, being involved in litigation, and indemnity cover. Multivariate models determined the adjusted odds ratios (aOR) and 95% confidence intervals (CI) after taking account of confounding; aOR > 1 indicated a positive association of factors with defensive practice while aOR < 1 indicated a negative association. There were 562 respondents. The most common defensive medical practice related to referring on a case after sensing the possibility of a complaint (365, 64.9%); the factors associated were senior grades (aOR 0.74, 95% CI 0.56–0.98, P = .04), private sector (aOR 1.27, 95% CI 1.008–1.61, P = .04), and indemnity cover (aOR 0.49, 95% CI 0.26–0.93, P = .03). The second most common defensive practice was calling inpatient admission, delaying discharge, additional testing, etc without medical indication and solely on patient or family request (265, 47.1%); the factors associated were age (aOR 0.46, 95% CI 0.33–0.64, P = .001), private sector (aOR 0.66, 95% CI 0.53–0.83, P = .001), and support by workplace staff (aOR 0.50, 95% CI 0.34–0.73, P = .001). Other defensive practices included refraining from difficult procedures or referring cases to another colleague due to the fear of complications (166, 29.5%) and unwillingness to accept patients in case of previous litigation history (157, 28.1%). This multivariable analysis in the United Arab Emirates found that higher age, higher job grades, indemnity cover and support by workplace staff reduced the odds of defensive medicine practice while working in the private sector had a mixed effect.

1. Introduction

Defensive medicine (DM) is the attitude adopted by medical professionals to avoid complex procedures or difficult patients and treating patients with their own legal safety in mind.^[1] Increasing number of medical lawsuits has forced medical professionals to adopt this behavior in order that they can remain free from or reduce any risk of litigation.^[2] This practice is becoming common worldwide according to the data reported from different countries including the Italy, Denmark, United Kingdom, Belgium, Egypt, and the United Arab Emirates.^[3–8] DM may increase the financial burden on the patients, delay treatment time and exposes patients to risks of unnecessary diagnostic interventions.^[9] According to a survey among American Academy of Orthopaedic Surgeons, DM is an important component between medical cost and patient benefits.^[10] Medical institutes are focusing attention on minimizing this practice by modifying practice rules and insurance policies, for example, the Belgian government has tried to provide better insurance policies to doctors by amending the laws.^[3] However, it has been reported in a review that tort reforms are not enough for health policies to control DM practices.^[11] Another review suggests other measures, for instance, improving the patient and physician communication, and awareness of medical-legal aspects to reduce the fear of medical misconduct to doctors for limiting increasing use of this practice.^[12]

There is a need to find specific issues or circumstances that eventually could lead to DM in the various behaviors of medical practitioners. A previously published Belgian study reported that the determinants of DM were gender, region, litigation experience, solo practice and an accessible incident reporting system.^[3] However, this study did not explicitly follow relevant reporting guidelines^[13] and, potentially, it addressed too many variables within a relatively small sample size risking overfitting of multivariate models.^[14] Our study focused on exploring the link between several factors that could be associated with different types of DM practices deploying multivariate analyses among a large sample of medical professionals in the United Arab Emirates.

2. Methodology

This is a multivariable logistic regression analysis of a survey on the practice of DM among medical professionals in the United Arab Emirates.^[5] The original study incorporated 23-point self-administered questionnaires and responses were taken from the medical practitioners working in Dubai and Abu Dhabi after taking ethical approval from Al Ain Medical District Human Research Ethics Committee (AAMDHREC). This study deployed recommended methods for logistic regression analysis^[14] to explore the association between various putative factors and various practices of DM. It is reported in compliance with STROBE.^[13]

Basic characteristics of the participants and details of their responses to the questionnaire items were reported in the previously published paper.^[5] For logistic regression analysis, DM was the outcome variable defined according to the responses given to questionnaire items about potentially unnecessary referral, testing, and additional care. DM behavior variables were defined as: referring on in case of sensing the possibility of a complaint; by doing following: calling inpatient admission, delaying discharge from inpatient, radiology/imagistic evaluation (ultrasound, MRI, CT, X-ray), laboratory tests, medication and repeated or unnecessary clinic appointment without medical indication and solely on patient or family request; refraining from difficult procedures or referring cases to another colleague due to the fear of complications; and unwillingness to accept patients in case of previous litigation history. For each of these we first gave them a binary coding and then ran the analyses to explore the determinants separately.

The regression analyses evaluated the factors: age, gender, medical specialty, job grade, years of practice, country of medical graduation, country of specialty board, current practice in hospital or private sector, feeling supported by workplace staff, being involved in litigation, and indemnity cover. These were explored as independent or predictor variables in univariate and multivariate models. Given the large number of independent variables, we respected the 10:1 event to variable ratio to avoid overfitting the models.^[13] Thus, if the responses to a questionnaire item capturing DM did not meet this criterion, we did not perform logistic regression. The model outputs generated adjusted odds ratios (aOR) and 95% confidence intervals (CI) after taking account of confounding; aOR > 1 indicated a positive association of factor with DM while aOR < 1 indicated a negative association. The analysis was carried out using SPSS version 21.

3. Results

There were 562 respondents with their basic age and gender distributions outlined in Table 1. Defensive behavior was exhibited by 28.1% to 64.9% of respondents depending on their response to questions about potentially unnecessary referral, testing, additional care, etc (Table 1). Only 31 (5.5%) responded affirmatively to the question about “referring difficult cases to colleagues even if competent to deal with them” and this was not considered further in multiple logistic regression due to small numbers.

Table 1.

Basic participants characteristics and binary outcomes variables used in logistic regression analysis to examine the determinants of defensive medical practice.

	Number	%
Basic characteristics Age (years)* 31–40 41–50 More than 50	126 174 252	22.1 31.0 44.8
Gender Female	158	28.1
Outcome variables
Referring in case of sensing the possibility of a complaint	365	64.9
Practicing the following: Calling inpatient admission, delaying discharge from inpatient, radiology/imagistic evaluation (ultrasound, MRI, CT, X-ray), laboratory tests, medication and repeated or unnecessary clinic appointment without medical indication and solely on patient or family request	265	47.1
Refraining from difficult procedure or referring case to another colleague due to the fear of complications	166	29.5
Unwilling to accept patient with previous litigation history	157	28.1

^* Some missing values not included in table.

Factors associated with practice of DM were evaluated for 4 specific questionnaire items that yield large percentage responses (Tables 2–5). With respect to referring in case of sensing the possibility of a complaint (Table 2), practiced by 365 (64.9%) respondents, there was a lower tendency among senior grades (aOR 0.74, 95% CI 0.56–0.98, P = .04), a higher tendency among those in the private sector (aOR 1.27, 95% CI 1.008–1.61, P = .04) and a lower tendency among those with indemnity cover (aOR 0.49, 95% CI 0.26–0.93, P = .03). With respect to calling inpatient admission, delaying discharge from inpatient, radiology/imagistic evaluation (ultrasound, MRI, CT, X-ray), laboratory tests, medication and repeated or unnecessary clinic appointment without medical indication and solely on patient or family request (Table 3), practiced by 265 (47.1%) respondents, there was a lower tendency among those with older age (aOR 0.46, 95% CI 0.33–0.64, P = .001), a lower tendency among those currently in private sector (aOR 0.66, 95% CI 0.53–0.83, P = .001), and a lower tendency among those who felt supported by workplace staff (aOR 0.50, 95% CI 0.34–0.73, P = .001). With respect to refraining from difficult procedures or referring cases to another colleague due to the fear of complications (Table 4), practiced by 166 (29.5%) respondents, there was a higher tendency among in those in private sector (aOR 1.30, 95% CI 1.01–1.67, P = .03), and a lower tendency among those who felt supported by workplace staff (aOR 0.53, 95% CI 0.36–0.79, P = .002). With respect to unwillingness to accept patients in case of previous litigation history (Table 5), practiced by 157 (28.1%) respondents, there was a lower tendency among those who felt supported by workplace staff (aOR 0.51, 95% CI 0.35–0.75, P = .001).

Table 2.

Logistic regression analysis to examine the determinants of defensive medical practice as measured by binary outcome: referring on in case of sensing the possibility of a complaint.

Determinant	Crude OR (95% CI)	P	Adjusted OR (95% CI)	P
Age	0.85 (0.69–1.04)	0.13	0.82 (0.60–1.12)	.23
Gender	1.55 (1.06–2.26)	0.02	1.42 (0.94–2.15)	.09
Medical specialty	0.72 (0.50–1.03)	0.07	0.70 (0.47–1.04)	.08
Job grade	0.75 (0.59–0.95)	0.01	0.74 (0.56–0.98)	.04
Years of practice	0.89 (0.67–1.18)	0.43	1.24 (0.80–1.92)	.31
Country of medical graduation	1.10 (0.74–1.62)	0.62	1.24 (0.77–1.98)	.36
Country of specialty board	1.17 (0.84–1.62)	0.33	1.21 (0.83–1.78)	.31
Current practice in hospital or private sector	1.15 (0.94–1.41)	0.16	1.27 (1.008–1.61)	.04
Feel supported by workplace staff	0.68 (0.47–0.96)	0.03	0.63 (0.43–0.92)	.01
Ever involved in litigation	1.07 (0.69–1.65)	0.74	1.16 (0.71–1.88)	.54
Indemnity cover	0.63 (0.35–1.13)	0.12	0.49 (0.26–0.93)	.03

Variable coding: age: increasing 10 years interval; gender: female = 1, male = 0; medical specialty: medicine = 0, surgery = 1; job grade: intern = 0, resident = 1, general physician = 2, specialist = 3, consultant = 4; years of practice: intern = 0, resident = 1, 0 to 5 years = 2, 5 to 15 years = 3, more than 15 years = 4; country of medical graduation: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1; country of specialty board: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1, not applicable if junior = 2; current practice; Govt. hospital/clinic = 0, both govt. and private hospital/clinic = 1, private hospital/clinic = 2; feel supported by workplace staff: yes = 1, no = 0; ever involved in litigation: yes = 1, no = 0; indemnity cover: none = 0, yes = 1.

CI = confidence interval, OR = odds ratio.

Table 5.

Logistic regression analysis to examine the determinants of defensive medical practice as measured by binary outcome: unwilling to accept patients in case of previous litigation history.

Determinant	Crude OR (95% CI)	P	Adjusted OR (95% CI)	P
Age	0.96 (0.78–1.18)	.71	0.85 (0.62–1.17)	.33
Gender	1.05 (0.71–1.56)	.77	0.89 (0.58–1.37)	.61
Medical specialty	1.22 (0.85–1.75)	.26	1.33 (0.90–1.98/0	.14
Job grade	0.80 (0.63–1.02)	.07	0.75 (0.56–1.00)	.05
Years of practice	1.05 (0.78–1.41)	.73	1.37 (0.88–2.14)	.16
Country of medical graduation	0.68 (0.45–1.03)	.07	0.72 (0.44–1.19)	.20
Country of specialty board	0.72 (0.51–1.006)	.05	0.81 (0.55–1.19)	.29
Current practice in hospital or private sector	1.11 (0.90–1.36)	.30	1.11 (0.88–1.40)	.37
Feel supported by workplace staff	0.59 (0.41–0.84)	.004	0.51 (0.35–0.75)	.001
Ever involved in litigation	1.58 (1.03–2.43)	.03	1.40 (0.87–2.26)	.16
Indemnity cover	1.22 (0.65–2.30)	.52	1.15 (0.58–2.27)	.68

Variable coding: age: increasing 10 years interval; gender: female = 1, male = 0; medical specialty: medicine = 0, surgery = 1; job grade: intern = 0, resident = 1, general physician = 2, specialist = 3, consultant = 4; years of practice: intern = 0, resident = 1, 0–5 years = 2, 5–15 years = 3, more than 15 years = 4; country of medical graduation: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1; country of specialty board: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1, not applicable if junior = 2; current practice; govt. hospital/clinic = 0, both govt. and private hospital/clinic = 1, private hospital/clinic = 2; feel supported by workplace staff: yes = 1, no = 0; ever involved in litigation: yes = 1, no = 0; indemnity cover: none = 0, yes = 1.

CI = confidence interval, OR = odds ratio.

Table 3.

Logistic regression analysis to examine the determinants of defensive medical practice as measured by binary outcome by doing following: calling inpatient admission, delaying discharge from inpatient, radiology/imagistic evaluation (ultrasound, MRI, CT, X-ray), laboratory tests, medication and repeated or unnecessary clinic appointment without medical indication and solely on patient or family request.

Determinant	Crude OR (95% CI)	P	Adjusted OR (95% CI)	P
Age	0.62 (0.50–0.76)	0.001	0.46 (0.33–0.64)	.001
Gender	1.23 (0.85–1.78)	0.26	1.12 (0.73–1.72)	.57
Medical specialty	0.78 (0.55–1.10)	0.16	0.77 (0.52–1.13)	.19
Job grade	0.89 (0.71–1.12)	0.35	0.93 (0.70–1.24)	.64
Years of practice	0.80 (0.60–1.06)	0.12	1.84 (1.18–2.87)	.007
Country of medical graduation	0.71 (0.48–1.03)	0.07	0.83 (0.52–1.33)	.45
Country of specialty board	0.95 (0.69–1.29)	0.74	0.97 (0.66–1.43)	.89
Current practice in hospital or private sector	0.60 (0.50–0.73)	0.001	0.66 (0.53–0.83)	.001
Feel supported by workplace staff	0.48 (0.34–0.68)	0.001	0.50 (0.34–0.73)	.001
Ever involved in litigation	1.35 (0.89–2.06)	0.15	1.61 (0.99–2.63)	.05
Indemnity cover	0.59 (0.33–1.07)	0.08	0.78 (0.40–1.53)	.48

Variable coding: age: increasing 10 years interval; gender: female = 1, male = 0; medical specialty: medicine = 0, surgery = 1; job grade: intern = 0, resident = 1, general physician = 2, specialist = 3, consultant = 4; years of practice: intern = 0, resident = 1, 0 to 5 years = 2, 5 to 15 years = 3, more than 15 years = 4; country of medical graduation: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1; country of specialty board: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1, not applicable if junior = 2; current practice; govt. hospital/clinic = 0, both govt. and private hospital/clinic = 1, private hospital/clinic = 2; feel supported by workplace staff: yes = 1, no = 0; ever involved in litigation: yes = 1, no = 0; indemnity cover: none = 0, yes = 1.

CI = confidence interval, OR = odds ratio.

Table 4.

Logistic regression analysis to examine the determinants of defensive medical practice as measured by the binary outcome: refraining from difficult procedures or referring cases to another colleague due to the fear of complications.

Determinant	Crude OR (95% CI)	P	Adjusted OR (95% CI)	P
Age	0.85 (0.68–1.05)	.14	0.93 (0.67–1.28)	.65
Gender	1.37 (0.92–2.03)	.11	1.13 (0.73–1.74)	.58
Medical specialty	0.98 (0.67–1.42)	.93	1.01 (0.67–1.52)	.95
Job grade	0.70 (0.55–0.89)	.005	0.73 (0.55–0.98)	.04
Years of practice	0.78 (0.58–1.05)	.10	0.97 (0.62–1.50)	.89
Country of medical graduation	0.92 (0.61–1.39)	.70	0.98 (0.59–1.61)	.94
Country of specialty board	0.95 (0.67–1.33)	.77	1.05 (0.70–1.56)	.82
Current practice in hospital or private sector	1.29 (1.04–1.61)	.01	1.30 (1.01–1.67)	.03
Feel supported by workplace staff	0.58 (0.40–0.84)	.004	0.53 (0.36–0.79)	.002
Ever involved in litigation	1.06 (0.67–1.66)	.79	1.12 (0.68–1.86)	.63
Indemnity cover	0.90 (0.48–1.69)	.75	0.83 (0.42–1.65)	.60

Variable coding: age: increasing 10 years interval; gender: female = 1, male = 0; medical specialty: medicine = 0, surgery = 1; job grade: intern = 0, resident = 1, general physician = 2, specialist = 3, consultant = 4; years of practice: intern = 0, resident = 1, 0 to 5 years = 2, 5 to 15 years = 3, more than 15 years = 4; country of medical graduation: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1; country of specialty board: UAE, other GCC countries, Asia, Africa = 0, USA, central and south America, Canada, European Union, other European countries, Australia, New Zealand = 1, not applicable if junior = 2; current practice; govt. hospital/clinic = 0, both govt. and private hospital/clinic = 1, private hospital/clinic = 2; feel supported by workplace staff: yes = 1, no = 0; ever involved in litigation: yes = 1, no = 0; indemnity cover: none = 0, yes = 1.

CI = confidence interval, OR = odds ratio.

4. Discussion

We aimed to delineate the possible factors that make a medical professional defensive towards his own legal safety. Our research showed that among the professionals employed in the United Arab Emirates the DM practice rates were a third to two-third depending on the responses to specific questionnaire items: The most common type of DM behavior was when a doctor noticed the risk of a patient complaint, and the least common was the avoidance of a challenging treatments out of concern for potential risks or declining to take patients with a history of medical lawsuits. Some factors increased the odds of DM while others decreased the odds. Higher practitioner age, higher job grades, indemnity cover and support by workplace staff significantly reduced the odds of DM. Working in the private sector had a mixed effect, increasing or decreasing the odds of DM depending on the specific DM practice.

The strength of our study is that we had a large sample size. This permitted a thorough exploration with rising overfitting in the multivariate model.^[14] We had over 50 more respondents than the previous study of associated factors.^[3] Odds ratios may be underestimated or overestimated if the event to variable ratio is low. We were careful not to exceed the 10:1 ratio recommended in the literature for avoiding overfitting the logistic regression models.^[14] Our approach used simple analysis before multivariate analysis as recommended.^[14] The univariate analysis is a akin to the chi square test in terms of examining relationship between the one variable and the outcome. The limitation of this study was a post hoc analysis of a survey conducted in 2 United Arab Emirates states, Dubai, and Abu Dhabi.^[5] The statistical output should be cautiously interpreted. It is acknowledged that exhibiting DM behavior in one or another form was subjected to multiple analyses taking responses to various questionnaire items. This approach may increase the possibility of spurious significance even though adjustment for confounding in multivariate analysis provides some protection against this risk. Thus, the P-values reported in our tables should not be taken to mean as if they are testing hypotheses; the analyses are exploratory with inferences useful for hypothesis generation. In this regard the sizes of association and the precision of CIs are more useful in interpretation.

The multivariate regression analysis explored responses from practitioners who varied in age groups and seniority levels. Older and senior practitioners practiced less defensively (aOR 0.46, 95% CI 0.33–0.64, P = .001, and aOR 0.74, 95% CI 0.56–0.98, P = .04, respectively for age and seniority). Junior-level professionals more frequently exhibited DM behavior (aOR 0.73, 95% CI 0.55–0.98, P = .04). Unlike the previous Belgian study,^[3] we did not find any association with gender across the DM practices subjected to multivariate analysis. Indemnity protection did reduce DM in our analysis. When recognizing patient complaints, the professionals in our study without indemnity coverage appeared to be acting defensively. Impression of support by colleagues also reduced DM. These factors appear to be consistent with previous findings. DM in common across both developed and developing regions.^[4,5,8]

The practitioners worked in various government and private clinics in the United Arab Emirates. It is interesting that working in the private sector had a mixed effect. It was associated positively with the DM behaviors referring on in case of sensing the possibility of a complaint and refraining from difficult procedures or referring cases to another colleague due to the fear of complications. However, it is associated negatively with the DM behavior whereby the respondents acknowledged that they call inpatient admissions, postpone inpatient discharge, perform laboratory tests, administer medication, schedule repeated or needless clinic appointments without a medical reason and only at the patient’s or family’s request, avoid difficult procedures due to potential complications, or refuse to accept patients who have previously been involved in medical litigation.

In conclusion, our study delineated the extent to which DM is likely to be linked with several factors through an analysis that adjusted for confounding. As DM is an evolving matter of concern in the medical field exploration of its determinants, as carried in our study, would be useful in determining how healthcare organizations can develop guidelines to bring it under control.

Author contributions

Conceptualization: Shamsa Al Awar.

Data curation: Shamsa Al Awar.

Formal analysis: Hassan Elbiss.

Methodology: Shamsa Al Awar.

Writing – original draft: Shamsa Al Awar, Hassan Elbiss.

Writing – review & editing: Gehan Sallam, Hassan Elbiss.