Skip to main content

Table 4 Univariate and multivariable logistic regression of the association between hospital caseload and treatment outcome

From: Towards better guidance on caseload thresholds to promote positive tuberculosis treatment outcomes: a cohort study

Main exposure/potential confounders/a priori confounders

Univariate regressiona

Multivariable regressiona

OR (95 % CI)

OR (95 % CI)

Hospital caseloadb

 

114+

P <0.001

e

 

73–113

1.13 (1.00–1.29)

 
 

27–72

1.06 (0.91–1.24)

 
 

<27

1.32 (1.13–1.55)

 

Notification date

 

Post-toolkit

P <0.001

P <0.001

 

Pre-toolkit

1.25 (1.17–1.33)

1.30 (1.21–1.40)

Location

 

Outside London

P <0.001

P = 0.16

 

Inside London

0.81 (0.71–0.93)

0.90 (0.77–1.04)

Sex

 

Male

P <0.001

P <0.001

 

Female

0.79 (0.75–0.84)

0.85 (0.80–0.90)

Age, years

 

<20

0.67 (0.60–0.75)

0.70 (0.63–0.79)

 

20–39

P <0.001

P <0.001

 

40–64

0.93 (0.87–0.99)

0.92 (0.85–0.98)

 

65+

1.48 (1.37–1.59)

1.44 (1.32–1.57)

Ethnic group

 

White

P <0.001

e

 

Black African

0.68 (0.63–0.75)

 
 

Black other

0.88 (0.75–1.03)

 
 

Indian subcontinent

0.76 (0.71–0.82)

 
 

Other

0.74 (0.67–0.82)

 

UK born

 

No

P <0.001

e

 

Yes

1.10 (1.03–1.17)

 

Site of diseasec

 

Extrapulmonary only

P <0.001

 

Pulmonary smear negative/unknown

1.16 (1.09–1.23)

 
 

Pulmonary smear positive

1.52 (1.42–1.63)

 

Drug sensitivity

 

Sensitive

P <0.001

 

Resistant to INH, EMB, or PZA

5.40 (4.86–6.00)

 
 

Not linked to a culture

0.87 (0.82–0.92)

 

Previous diagnosis

 

No

P <0.001

 

Yes

1.32 (1.19–1.46)

 

Social risk factorsd

 

No or unknown

P <0.001

P <0.001

 

One or more previous

1.33 (1.11–1.60)

1.36 (1.12–1.66)

 

One or more current

2.83 (2.44–3.28)

2.67 (2.28–3.13)

  1. Univariate and multivariable random effects logistic regression of the association between hospital caseload and treatment outcome, England, 2006–2012. All regression models adjusted for clustering by hospital; 42,995 cases in the multivariable model, dashed variables not retained
  2. aOdds of an unfavourable versus a good or neutral treatment outcome; records missing an outcome excluded
  3. bMean caseload per hospital over the preceding 3 years
  4. cPulmonary site could be with or without extrapulmonary disease
  5. dSocial risk factors a composite variable of homelessness, imprisonment, drug misuse, and alcohol abuse; current risks override previous risks
  6. eInteraction present between hospital caseload, ethnicity and UK born, see Table 5
  7. CI Confidence interval, EMB Ethambutol, INH Isoniazid, OR Cluster-specific odds ratio, p P-value, PZA Pyrazinamide