Table 4.

Summary of results: additional information

Study and country	N at baseline	Intervention and outcome measure	Level at baseline	Level postintervention		Effect on outcome
Randomised controlled study designs
Baker et al23 UK	Practices: 17 (I) 16 (C)	Intervention(s): G+FB Measure(s): median (IQR) number of TFTs per 1000 registered patients	17.4 (8.0, 39.5) (I) 22.7 (10.4, 30.9) (C)	19.8 (6.2, 42.3) (I) at 6 months 19.5 (10.3, 31.1) (C) at 6 months 17.7 (6.6, 43.3) (I) at 9 months 17.3 (10.1, 34.0) (C) at 9 months 13.2 (6.3, 35.7) (I) at 12 months 20.9 (13.3, 35.3) (C) at 12 months		For TFTs, the difference in mean change in test rate from baseline to 4th quarter was—1.45 in favour of the I (95% CI −4.59 to 1.68) but was not statistically significant (p=0.35) The intervention had no significant effect on the other tests, too.
Daucourt et al,25 France	Hospital wards: 17 (PMC+TRF) 20 (TRF) 17 (MPC) 13 (C)	Intervention(s): PMC, TRF, PMC+TRF Measure(s): GCR	NA	77.9% (95% CI 68.9% to 87.0%) (MPC+TRF) 82.6% (95% CI 73.1% to 92.1%) (TRF) 73.4% (95% CI 56.7% to 90.1%) (MPC) 62.0% (95% CI 47.7% to 76.4%) (C)		GCR was significantly higher in PMC+TRF compared with the C (OR 2.65; 95% CI 1.52 to 4.62; p<0.01); slightly lower compared with TRF and slightly higher compared with MPC but the differences were not statistically significant No difference between MPC and the C (OR 1.28; 95% CI 0.75 to 2.19; p=0.37)
Thomas et al,10 UK	Practices: 21 (FB+R) 22 (FB) 22 (R) 20 (C)	Intervention(s): FB, R, FB+R Measure(s): median (IQR) TFT requests per 10 000 patients per practice	750 (515–1329) (C) 829 (476–1412) (FB) 961 (476–1338) (R) 891 (392–1277) (FB+R)	795 (552–1466) (C) 802 (432–1359) (FB) 891 (490–1250) (R) 800 (287–1077) (FB+R)		Is were significantly less likely to request TFTs (FB group: OR 0.90 (0.84 to 0.97), p=0.005; R group: OR 0.82 (0.83–0.95), p=0.001). Across all targeted tests, intervention practices were significantly less likely to order tests.
Tierney et al,35 USA	Scheduled visits: 7658 (I) 7590 (C)	Intervention(s): display of computer-generated probability estimates Measure(s): charges per scheduled visit (in USA$)	NA	1.25 (C) 1.12 (I)		TSH showed 10.3% decrease in charges per visit in the I group but this difference was not significant at p=0.05. Across all targeted tests, there was a significant reduction in charges per visit (−8.8%, p<0.05), with return to baseline at 3 months follow-up.
Non-randomised controlled study designs
Berwick and Coltin,17 USA	35 internists and 30 adult nurses at 3 centres (total number of visits not given)	Intervention(s): TSE, FBC and FBY Measure(s): per cent change in the number of TT4 ordered per 100 encounters and the CV of test ordering rates	TT4 tests per 100 encounters: Site X: 72 Site Y: 72 Site Z: 45 Range of rates of TT4 use: Site X: 40–273 Site Y: 36–122 Site Z: 35–78	Change:	CV:	TT4 use declined in the TSE and FBC groups but increased in FBY; CV decreased in FBC and FBY but increased in the C group and showed very small increase in TSE group. The statistical significance of these results is NR. Across all tests statistically significant decline in test ordering (14.2%, p=0.012) was observed only in the FBC group.
				C: +1.7%	+15.7%
				TSE: −15.9%	+0.5%
				FBC: −12.1%	−23.6
				FBY: +34.0%	−28.2
Gama et al,18 UK	Outpatient visits: 2991 (I) 4393 (C)	Intervention(s): FB Measure(s): TFTs per outpatient visit (mean, range)	0.17 (0.12–0.22) (I) 0.05 (0.04–0.06) (C)	0.13 (0.11–0.17) (I) 0.06 (0.05–0.08) (C)		The mean number of TFTs per outpatient visit decreased by 21.9% in the I group (p<0.01) and increase by 20.8% (not significant) in the C group. Similar results were obtained for the other tests for outpatients but not for inpatients (data NR).
Schectman et al,33 USA	1425 patients, 30 clinicians (group distribution not given)	Intervention(s): R and R+FB Measure(s): Compliance Mean (SE) number of TFTs ordered per patient	68% (R) 65% (R+FB) 1.68 (0.04)	81% (R) at 6 months 77% (R) at 12 months 64% (R+FB) at 6 months 80% (R+FB) at 12 months 1.37 (0.03) at 2/12 (following EM) 1.32 (0.05) at 6-month follow-up 1.49 (0.04) at 1-year follow-up		Significant increase in compliance in the reminder group (p=0.05) but not in the R+FB group; however, after excluding an outlier both groups had similar increase in compliance (77% vs 80%, p=0.39). The mean number of TFTs ordered per patient also decreased significantly (no p value given) but increased again at 1-year follow-up (only data combining both groups provided). TSH levels increased significantly while TT4 and T3RU decreased but no details are given.
Tomlin et al,36 New Zealand	GPs: 3140 (I) 2443 (C)	Intervention(s): E+G+FB Measure(s): Tests per year per GP: TSH: FT4: FT3: Total number of TFTs: Ratios of different TFTs: TSH/FT4 TSH/FT3 Expenditure (%)	223.6 (I) 33.8 (C) 144.2 (I) 29.1 (C) 41.6 (I) 11.0 (I) NR 2.4:1 7.1:1 NR	215.2 (I) 32.0 (C) 80.7 (I) 25.3 (C) 26.6 (I) 9.4 (C) 21% decrease (I) NR (C) 3.0:1 8.5:1 −19.8% (I) −9.5% (C)		TSH showed small decrease (4%, p<0.01) in the I group and no change in the C group (p<0.11). FT4 and FT3 decreased by 44.1% and 36.0%, respectively (p<0.01) in the I group and 13.1% and 14.6%, respectively, in the C group (p<0.01). In the I group, the proportion of TSH as the sole test ordered increased from 43.2% to 65.2% (p<0.001). Ratios of TSH to FT4 increased from 2.4:1 to 3.0:1 and TSH to FT3 from 7.7:1 to 8.5:1. Simultaneous testing of TSH and FT4 and/or FT3 decreased by 41.1% and there was a decrease in the net TFT expenditure (no p values given).
Wong et al,42 USA	NR	Intervention(s): G+TRF Measure(s): tests per month	Intervention tests (months to intervention):* T4 (RIA) and T3RU: 9 months : 1100 6 months: 1150 3 months: 1100 TSH: 9 months: 900 6 months: 1300 3 months: 900 T3 (RIA): 9 months: 950 6 months: 1000 3 months: 900 Control tests:* CK: 9 months: 1000 6 months: 980 3 months: 980 LDH: 9 months: 650 6 months: 700 3 months: 750	Intervention tests (months after intervention was introduced):* T4 (RIA) and T3RU: At 2 months: 1000 4 months: 1000 6 months:1100 8 months:950 TSH: 2 months: 500 4 months: 500 6 months:600 8 months:500 T3 (RIA): 2 months: 200 4 months: 300 6 months: 400 8 months: 300 Control tests*: CK: 2 months: 1100 4 months: 700 6 months:1000 LDH: 2 months: 700 4 months: 500 6 months:700		Distributing guidelines through a bulletin alone failed to produce effect but in combination with request form redesign it led to restructuring of test ordering patterns with decrease of ‘complete’ thyroid panel and increase of hyperthyroid and hypothyroid panels and thyroid function screen. No changes in T4 (RIA) and T3RU but the number of T3 (RIA) and TSH tests ordered per month fell on the average to 38% and 61%, respectively, of the mean monthly rates at which these tests had been ordered in the preceding 18 months. No changes were observed in the ordering of the control tests. Statistical significance of the above results is NR. Not all data presented here!
Interrupted time series
Horn et al,45	Average monthly orders per 1000 patient visits (TSH): 174.1 (I) 140.3 (C)	Intervention: display of cost of tests being ordered Measure(s): comparison of change-in-slope of the monthly ordering rates between intervention and control physicians for 12 months preintervention and 6 months postintervention	Per cent change in monthly order rates (TSH, preintervention): 0.2% (I) 0.1% (C)	Per cent change in monthly order rates (TSH, postntervention): 0.5% decrease (I) 0.4% increase (C)		The difference in the rate of change preintervention to postintervention was 0.7% decrease in the I group and 0.3% increase in the C group; none of these results was significant at p value <0.002 (2-sided Bonferroni-adjusted p value). Across all 27 evaluated tests, a statistically significant modest decrease in ordering rates of intervention physicians compared with control physicians was observed in 5 tests.
Van Walraven et al,39 Canada	NR	Intervention(s): G+CFP; G+TRF Measure(s): avoided tests, utilisation rate and cost		1 July 1991: TSH: 1 per 100 persons* TT4: 1.2 per 100 persons*	G+CFP: TT4: 4359 (95% CI −14 to 23 430) tests avoided T3RU: 3073 (−28 to 18 153) tests avoided G+TRF: TSH: 2200 (95% CI −1638 to 6039) tests avoided	G+CFP led to 96% decrease in the TT4 utilisation (p=0.02) and decrease in T3RU. Guidelines plus removing TSH ‘tick box’ from the request form resulted in 12% decrease in TSH utilisation (p=0.03). All interventions together resulted in a decrease of 626 098 tests, which saved $2 010 400, including $29 664 in the final year.
Before and after study designs
Adlan et al,22 UK	Admissions: 1593 (pre) 1176 (post)	Intervention(s): G Measure(s): proportion of admitted patients offered TFTs	53.8% (857 out of 1593)	21.7% (255 out of 1176)		Significant reduction (32.1%, p<0.001) in the proportion of admitted patients offered TFTs
Chu et al,24 Australia	ED visits: 24 652 (pre) 25 576 (post)	Intervention(s): G Measure(s): mean number of TFTs ordered per 100 ED visits	2.2 (20-week preintervention period)	1.6 (20-week postintervention period)		Significant reduction in the mean number of TFTs (0.6 tests per 100 ED presentations, p=0.001) The mean number of all blood tests ordered per 100 ED presentations fell by 19% (p=0.001) and the mean cost fell by 17% (p=0.001).
Cipullo and Mostoufizadeh,39 USA	NA	Intervention(s): G Measure(s): mean tests utilisation (T3 per discharge)	0.006 (1 year before)	0.005 (1 year after)		The number of T3 uptake per discharge decreased by 17%. Most of the other targeted tests also showed decline in utilisation. Statistical significance NR.
Dowling et al,26 USA	Patient visits: 10 961 (pre), 6606 (post),3024 (at 5 months follow-up)	Intervention(s): E+FB Measure(s): proportion of indicated TSH (out of all TSHs performed) Rate of TSHs ordered per patient visit (total number of TSHs and visits) Rate of indicated TSH per visit Rate of non-indicated TSHs per visit	28% (25 of 90) 1 per 122 (90 in 10 961) 1 per 438 1 per 169	65% (15 of 23) 43% (9 of 21) (at 5 months follow-up) 1 per 287 (23 in 6 606) 1 per 178 (21 in 3 024) (at 5 months follow-up) 1 per 440 1 per 336 (at 5 months follow-up) 1 per 826 1 per 252 (at 5 months follow-up)		The proportion of indicated TSHs increased significantly (p<0.001) while TSHs per patient visit decreased significantly (p<0.0001) in the intervention period but both showed some decline at 5 months follow-up. The rate of indicated TSHs per visit did not change significantly while the rate of non-indicated TSHs per visit decreased drastically in the intervention period but increased again at follow-up. Data for the control test, CBC with differential, is not shown here but the rate of ordering showed steady decline even in the follow-up and the rates of ordering both indicated and non-indicated CBCs decreased at the end of the intervention, although the statistical significance of these results was NR.
Emerson and Emerson,27 USA	Unclear	Intervention(s): TRF Measure(s): number of tests/panels ordered preintervention to postintervention	TSH: 5300* FT4: 750 TT4: 1700 TT3: 800 FTI/T3RU: 900 TFT cascade: NA Combined TSH and cascade: 5250	TSH: 3000* FT4: 1450 TT4: 200 TT3: 500 FTI/T3RU: 100 TFT cascade: 1700 Combined TSH and cascade: 4750		TFT testing decreased significantly (p<0.01) with a shift towards FT4 and thyroid cascade. Across all tests, the total number of tests remained the same (due to an increase in the number of patients) but the number of tests per patient visit showed significant decrease (p<0.01).
Feldkamp and Carey,28 USA	Sequential TFT requests: 1000 (pre) 463 (post, 1 year) 625 (post, 3 years)	Intervention(s): G Measure(s): percentage of different TFTs and combinations:	Prealgorithm:	1 year	2 years	‘TSH only’ and DRTSH accounted for 92.4% of all TFTs 3 years after the introduction of the algorithm. The other combinations gradually decreased. However, the statistical significance of these results is NR.
		TFT only	33.3%	44.8%	32.2%
		DRTSH algorithm	–	24.4%	60.2%
		TSH+TT4+T3RU	16.6%	3.9%	1.3%
		TT4 only	10.0%	4.8%	0.8%
		TT4+T3RU	6.8%	2.8%	1.1%
		Others (including TT3)	–	–	3.0%
		TFTs per 1000 patients:	Prealgorithm:	Postalgorithm: DRTSH:	Tests/1000
		TSH	832	982	1000
		TT4	667	216	202
		T3RU	234	159	202
		Total	1 733	1359	1404
		Difference	–	374	329
Grivell et al,29 Australia	NR	Intervention(s): FB Measure(s): ratio of thyroxin requests postintervention to preintervention	NA	1.20		Thyroxin requests in the postintervention period were 1.2 times the requests in the preintervention period but the statistical significance of this result was NR.
Hardwick et al,44 Canada	NR	Intervention(s): G+CFP Measure(s): proportion (number) of different TFTs:	1974/75	1976/77	1978/79	Overall decline from baseline to 3 years postintervention (1978/79) with shift towards TT4 which accounted for 80.4% of all TFT investigations in the last period. Expenditure also decreased to 34% of the expected charges by the end of the study period. The statistical significance of these results was NR.
		TT3	21.8% (29 004)	19.0% (35 101)	4.7% (7502)
		TT4	51.8% (68 912)	50.8% (93 988)	80.4% (128 343)
		ETR	26.4% (35 183)	30.2% (55 798)	14.9% (23 703)
		Total	100% (133 099)	100% (184 887)	100% (159 548)
			1975/76	1977/78
		TT3	20.4% (33 334)	11.8% (19 255)
		TT4	49.6% (81 004)	62.3% (101 805)
		ETR	29.8% (48 832)	25.9% (42 346)
		Total	100% (163 170)	100% (163 406)
Horn et al,45	Physicians: 153 (I) 62 (C)	Intervention(s): Display of cost of tests being ordered (computer-based ordering system) Measure(s): difference in per cent change in monthly orders between I and C group (orders per 1000 patient visits)	Baseline average monthly order rate (orders per 1000 visits): 174.1 (I) 140.3 (C) Per cent change in monthly order rate: 0.2% (I) 0.1% (C)	Per cent change in monthly order rate: −0.5% (I) 0.4% (C)	Difference: −0.7% (I) 0.3% (C)	Monthly order rates for TSH decreased slightly in the I group and increased in the C group but the difference was not statistically significant (p=0.04; because 27 tests were analysed, the study used a 2-sided Bonferroni-adjusted p value of <0.002 to determine statistical significance).
Larsson et al,30 Sweden	19 primary care centres	Intervention(s): E Measure(s): mean ratios of the requests for related tests:	1996	1997		Significant decrease in TT3/TSH (difference between mean ratios 0.073, SD=0.089, p=0.0012) and non-significant decrease in FT4/TSH (difference 0.032, SD=0.116, p=0.13). As recommended, the ratio of TSH to all TFTs increase significantly (difference −0.017, SD=0.041, p=0.048).
		TT3/TSH	0.129	0.056
		FT4/TSH	0.333	0.301
		TSH/total number of TFTs	0.124	0.141
Mindemarkand Larsson31 (follow-up of Larsson 1999)		Median ratios:	1997	2004		7 years after the intervention TT3/TSH and (TT4+FT4)/TSH were not significantly different from those at the end of the original study period, thus showing no decay in the intervention's effect. However, THS/all TFTs showed slight decrease (difference −0.043) which was statistically significant (p<0.05). Most of the other tests’ ratios also remained stable.
		TT3/TSH	0.029	0.022
		(TT4+FT4)/TSH	0.273	0.237
		TSH/all TFTs	0.115	0.072
Nightingale et al,32 UK	Number of patients: 654 (before) 833 (after)	Intervention(s): PMS+E+FB Measure(s): change in compliance	55%*	85%*		Compliance of TSH requests increased but results are given only as a graph and the statistical significance is NR. Across all tests, the total number of tests requested per patient day declined by 17% (p<0.001).
Rhyne and Gehlbach,43 Canada	NR	Intervention(s): G+E Measure(s): TFPs per 100 encounters Proportion of: ‘high indications’ ‘low indications’	October to December 1976 1.00 January to March 1977 Approximately 1.10 45% 29%	June to August 1977 Approximately 0.70 September to November 1977 Approximately 1.00 53% 19%		Significant decrease in the number of TFP ordered per encounter in the 3 months after the intervention (p<0.05) but return to baseline in the following 3 months. Results given only as a graph. The proportion of FTP for ‘high indications’ increased and that for ‘low indications’ decreased but was not statistically significant (p<0.05). Senior residents decreased their TFP ordering rate while that of junior residents increased (p<0.05).
Schectman et al,33 USA	1425 patients, 30 clinicians (group distribution not given)	Intervention(s): EM Measure(s): compliance rate Mean (SE) number of TFTs ordered per patient	35% 1.68 (0.04)	67% at 2 months (following EM) 1.37 (0.03) at 2 months (following EM) 1.49 (0.04) at 1 year follow-up		Significant increase in compliance after EM (p<0.0001) The mean number of TFTs ordered per patient decreased significantly following EM (p<0.0001) and showed further decline at 6 months after the subsequent interventions (see under Non-randomised controlled studies above) but increased again at 1 year follow-up. The educational intervention had greater impact on nurses and physician assistants than physicians (absolute increase in compliance 63% vs 28%).
Stuart et al,34 Australia	NR; annual census of 42 500 patients	Intervention(s): E+G+FB Measure(s): mean cost of TFT (in $A) per patient	0.426	0.047		TFT ordering decreased by 89% and showed a significant decrease in cost per patient (−89% difference (95% CI −55% to −123%; p<0.001). Across all tests, there was 40% decrease in the ordering of tests with test utilisation falling from a mean of $39.32/patient to $23.72/patient (p<0.001). Tests not allowed to be ordered for ED patients, such as TFT, showed the greatest decrease. The effect was sustained at 18 months follow-up.
Toubert et al,37 France	800-bed hospital with annual census of 25 266 inpatients and 242 013 outpatients	Intervention(s): G+R Measure(s): number of TFTs:	1996	1997	1998	A substantial decrease in the total number of TFTs mainly due to a decrease in the number of FT3 and FT4; a decrease in the relative proportion of FT3. Single TSH order forms increased from 23% to 50%, while TSH+FT4+FT3 decreased. The statistical significance of these results was NR. The percentage of appropriate tests increased from 42.5% (95%CI 39.9% to 45.1%) in 1996 to 72.4% (95%CI 69.8% to 75%) in 1997 (p<0.0001) but there was some decrease in 1998 (no p value given). (Data for thyroid antibodies is not presented here.)
		Total TSH tests	1329	1119	1062
		Total FT4 tests	1011	535	539
		Total FT3 tests	715	247	226
		Total number of TFTs	3145	1901	1827
		Patterns:
		Single TSH	305	563	512
		TSH+FT4	319	313	333
		TSH+FT3	23	25	20
		TSH+FT4+FT3	682	218	197
		FT4+FT3	10	4	9
		All TFT request forms	1339	1123	1071
		Appropriateness:	42.5%	72.4%	70.7%
van Gend et al,38 The Netherlands	NR	Intervention(s): TRF+FB Measure(s): ratio of FT4 (removed from) to TSH (left on the request form)	1992 0.96 (2498:2608)	1993 0.31 (1180:3747) 1994 0.28 (1436:5040)		There was a decrease in the FT4:TSH ratio indicating that the intervention had impact on test ordering patterns but the statistical significance of the results was NR.
Vidal-Trécan et al,40 France	June 1995: all TFTs: 27 945	Intervention(s): E+G+TRF Measure(s): number (%) of TFTs	June 1995	June 1998		11% decrease in the total number of TFTs even though the number of admissions increased by 2% and the number of outpatient visits by 6%. The proportion of FT4 tests remained the same (33%); the proportion of FT3 measurements decreased by 6% and the proportion of TSH tests increased. The statistical significance of the results was NR.
		Total TFTs	100% (27 945)	88.7% (24 794)
		FT3	20% (5491)	14% (3534 of 24 794)
		TT3	1% (339)	1% (371 of 24 794)
		FT4	33% (9301)	33% (8125 of 24 794)
		TT4	2% (478)	1% (238 of 24 794)
		TSH	44% (12 336)	51% (12 526 of 24 794)
Willis and Datta,41 UK	An average of 950 patients and 309 thyroid profiles per month	Intervention(s): E+G Measure(s): mean (SD) number of TFT profiles per admission	0.32 (0.05)	0.08 (0.01)		Significant decrease in the number of requested thyroid profiles (p<0.001). Across all tests, a significant change between the total number of sets requested per admission before (7.5 (0.87)) and after the intervention (5.9 (0.33)), p<0.001.

*Approximate reading off a graph.

C, control group; CBC, complete blood count; CFP, changes to funding policy; CK, creatinine kinase; CV, coefficient of variation; DRTSH, directed thyroid testing algorithm; E, education; ED,emergency department; EM, educational memorandum; ETR, effective thyroxin ratio; FB, feedback; FBC, feedback on cost; FBY, feedback on yield; FT3, free tri-iodothyronine; FT4, free thyroxine; FTI, free thyroid index; G, guidelines/protocol; GCR, guideline conformity rate, the proportion of TFTs ordered in accordance with the guidelines; GP, general practitioner; I, Intervention group; LDH, lactic dehydrogenase isoenzyme; NA, not available; NR, not reported; PMC, pocket memory card; PMS, protocol management system; R, reminders; T3 (RIA), tri-iodothyronine radioimmunoassay; T3RU, tri-iodothyronine resin uptake; T4 (RIA), thyroxin radioimmunoassay; TFP, thyroid function panel; TFTs, thyroid function tests; TRF, test request form redesign; TSE, test-specific education; TSH, thyroid-stimulating hormone; TT3, total tri-iodothyronine; TT4, total thyroxin.