Table 3.

Source Interventions (Type A).

Authors	Intervention & Study		N, Response Rate & Method	Exposure Levels		Change in Levels and Distribution of Change across Participants	Outcome Measure(s) before and after Outcomes	Did Outcome Change with Change in Exposure? Yes/No (Significance Tested?)	Before/after Outcome Change Compared to That Estimated from an ERF	Comments	Confounders Adjusted for in Analyses
	Nature	Design		Before	After
Brown (2015) [17]	Brisbane, Truck restriction, change in traffic composition Note: the date of this paper was outside the search time period	B/A. Five rounds of after surveys out to 20 months	99 in panel Response rate 84% ~20% of panel drop out each survey round Interviews	65–73 Lden 60–68 Lnight 69–77 L10,18 h Measured	65–73 Lden 60–68 Lnight 69–77 L10,18 h Measured	No change in Lden, Lnight or L10,18 h But see comments All Ps experienced same change—but were exposed to different before levels	%HA based on 7, 8 & 9 of ISO (but with 0–9 scale). B: 58% HA A: 33%, 18%, 18% HA respectively at survey rounds 2, 3 & 4 Mean Annoyance also reported	n.a. as no change in Lden exposure (but there was a change in number of noise events) Est. Marg. Mean annoyance scores changed significantly over period of truck restriction (F4,170.4 = 12.18, p < 0.001) (see comments)	ERF cited was Miedema & Oudshoorn (2001) [13] 58% HA in before-study much higher than estimated by ERF (latter is 16 to 30% for Lden over the range of Ps’ exposures No observation possible on the relationship of change in outcomes with the ERF because Lden did not change	Change in response attributed to change in number of noise events	Noise sensitivity; neighborhood quality; respondent association with trucking industry.
Pedersen, Le Ray, Bendtsen & Kragh (2013) [18]	Copenhagen Resurfacing with noise reducing pavement.	B/A study 12 mo. After Not repeated measure	2870 over two areas near roads Response rate 41% Mail surveys	42–74 Lden Modelled noise map. Note: wide range of before levels	38–70 Lden	Measured 4 dB reduction in source levels Same reduction assumed for all Ps	%HA based on 8, 9 & 10 of ISO (0–10 scale) Mean Annoyance also reported	Yes B&A mean annoyance scores were different (Welch’s t-test, p < 0.001)	Authors reported logistic regression ERFs for each of before and after conditions (n = 2870). The 95% CIs of B & A curves tended to overlap, and authors merged the data to establish the ERF. Hence change in response to −4 dB intervention estimated by the ERF. B & A ERFs curves are overlapping—largely parallel but with ERF (after) slightly lower than ERF (before). Response to change estimated by ERF. Slightly lower ERF(A) indicates excess response. The authors also report ‘…a small tendency to a lower %HA in the 50–60 dB range in the after situation…’.	Merged ERF was higher than Miedema & Oudshoorn (2001) [13] ERF over 60–74 Lden
Stansfeld, Haines, Berry & Burr (2009) [19]	UK Bypass roads constructed reducing traffic flow in three small towns	B/A study B:1 year A: 6–7 mos	17 5 exposed 184 control Response rate B:70% A: 74% 67 Ps at exposed area follow-up Delivered questionnaire	L10,3 h (& Leq,3 h) Exposed: 75–78 Control: 55–58 Measured Includes train noise	See next column	Change in L10,3 h of −2 to −4 dB suggested for most locations No reporting of distribution of these small changes across Ps.	‘Standard’ noise question for assessing level of annoyance with environmental noise at home. No significant change in mean annoyance score with intervention.	No change in annoyance. Explanation was that the change was too small to be noticed	n.a.	Changes in traffic flow on source roads were small: 24 h flow changed from 26 k to 23 k veh/day, and 24 k to 21 k veh/day
Baughan & Huddart (1993) [20]	U.K. Decreased traffic flow at 14 sites; increased traffic flow at 6 sites; 2 control sites	B/A study + controls 1–2 mos B&A changes Repeated measure	33–50 per site Response rate and dropout rate not reported Interviews	L10,18 h Decrease sites: 66–76 Increase sites: 65–78		L10,18 h 14 sites with changes ranging from −15 to +5 dB	7 point numerical scale of satisfaction with level of road traffic noise with endpoints labelled Def. Satis. And Def. Unsatis Outcome reported as mean dissatisfaction score	Yes Infer from next column No statistical tests reported	Authors refer to ERF derived from ‘TRRL’ survey at 35 steady-state sites. Authors conclude: For decreases, both before and after levels (of dissatisfaction) differed significantly from steady state (ERF). B/A transitions steeper than ERF; For increases, after levels differed significantly from steady state. B/A transitions steeper than ERF; No statistical tests reported; Response to change in same direction as estimated by ERF, but much steeper, indicating excess response	Data used in Griffiths & Raw (1989) [21] below also included in analysis in this paper
Griffiths & Raw (1989) [21]	England. Repeated measure of after survey in Langdon & Griffiths (1982) [22] 5 sites	Repeat of After study at 7–9 years. After 48% of Ps repeat interview	430 Interviews	See Langdon & Griffiths (1982) [22]			Four-point verbal bother scale Outcome reported as mean bother score for each of B&A conditions	n.a.(because there was no change in exposure between 7 and 9 years)	See results in Langdon & Griffiths (1982) [22] below. Observed Excess responses show no diminution out to 2 years after change, but is diminished, but still exists, 7–9 years after the change
Brown (1987) [23]	Brisbane. Increase in traffic flow	B/A study 2 weeks B, 7 & 19 mos A Repeated measure	20 Response rate 83% Interviews	LAeq,24 h 60 L10,18 h 60 Ldn 61	LAeq,24 h 66/67 L10,18 h 68/71 Ldn 69/71	LAeq,24 h + 6/+7 dB L10,18 h +8/+11 dB Ldn +8/+10 dB	7 point semantically labelled annoyance scale. Reported individual responses and %HA based on top two categories.	Yes Distribution of individual annoyance responses changed after intervention (Friedman Two-way Anova, p < 0.01). 90% CIs for %HA B & A intervention do not overlap	ERF cited was Schultz (1978) [10] and plotted as band containing 90% of data points used in Schultz synthesis. Before %HA lay within Schultz 90% band, After %HA lay above ERF (though Cis for %HA are wide due to small sample size). Indicates excess response to increase in exposure	Note: No evidence of adaptation. Distribution of annoyance scores not different at 7 and 19 mos after change (t-test, p < 0.05)
Griffiths & Raw (1986) [24]	England. Decreased traffic flow at 6 sites; increased traffic flow at 2 sites	B/A study 1–4 mos before change 2–3 mos after change Repeated measure	469 Response rate 74% 17% drop out between surveys (391)	L10,18 h Decrease sites: 65–81 Increase sites: 54–56 Measured and calculated	L10,18 h Decrease sites: 66–74 Increase sites: 61–69	Change in L10,18 h at seven sites were: (1) −14.5 dB (2) −5.7 dB (3) −2.6 dB (4) −3.1 dB (5) −1.3 dB (6) +5 dB (7) +15 dB	7 point numerical scale of satisfaction with level of road traffic noise with endpoints labelled Def. Satis. And Def. Unsatis Reported site mean dissatisfaction scores	Yes Infer difference between B&A mean dissatisfaction scores from next column	Authors calculated ERF using ‘steady-state’ before responses. Site mean dissatisfaction scores regressed against before L10,18 h. Mean dissatisfaction scores (After) were compared to those estimated by the ERF. For decreases: decrease in site mean dissatisfaction score was greater than estimated by a conservatively estimated ERF (t = 3.14, df = 4, p < 0.025). Similarly, at two increase sites, increase in individual dissatisfaction score was greater than estimated (t = 2.93, df = 81, p < 0.005). Response to decrease/increase changes in direction estimated by ERF, but steeper—hence excess response	Note: resurvey of three decrease-sites out to 17–22 mos. After change showed no change in observed excess response. Griffiths & Raw (1989) [21]
Brown, Hall & Kyle-Little (1985) [25]	Brisbane. Reduction in traffic flow	B/A study with 2 control groups (quasi experimental)	49 Cntrls: 52, 40	L10,12 h 74.3 Calculated Cntrl. 75.1 Measured	L10,12 h 64.5 Measured Cntrl. 65.2 Measured	L10,12 h −9.8 dB	7 point semantically labelled annoyance scale. Reported %HA based on top two categories. Annoyance with before conditions assessed in retrospect.	Yes Note, before %HA based on retrospective assessment No statistical test reported—but see next column	ERF cited was Schultz [10] and plotted as band containing 90% of data points used in Schultz synthesis. After %HA and Control sites %HA lay within 90% band. Before %HA (retrospective) lay outside 90% band. Response to decrease in same direction as estimated by ERF, but steeper, indicating excess response	This study relied on retrospective assessment of annoyance before the change
Langdon & Griffiths (1982) [22]	U.K. Reduction in traffic volumes at 6 sites after opening of new relief roads	B/A study 2–3 mos. B to 4–6 mos. A	Number of respondents at each of the six sites not reported	LAeq,24 h 72–76.5	LAeq,24 h 56.5–73.5	Change in LAeq,24 h 6 sites ranging from −5 to −15.5 dB All Ps at a site experienced the same change in exposure	Four-point verbal bother scale Outcome reported as median of bother score for each of B&A conditions at all six change sites	Yes Infer difference between B&A median bother scores from next column	Authors compare change data to ERF from eight sites in London where exposure and response were measured under steady-state condition. For the six change sites: No sig. diff. between median Before bother scores and scores estimated by ERF (t-test = 2.13, p > 0.05). Sig. diff between median After bother scores and scores estimated by ERF (t-test = 8.25, p < 0.001) Response to change in same direction as estimated by ERF, but steeper, indicating excess response
Kastka (1981) [26]	Germany Complex set of noise and traffic control measures in 6 cities Plots 50 data points	B/A study	1800	Measured Ld (range 47–68 B, 50–65 A		Range of sites with changes of −8 to +3 dB (mean −1 dB)	Complex set of measures including assessment of sensory experience and interferences of noise	Yes Infer from next column No statistical tests reported Percentage highly annoyed in line with an extra 6 dB from steady-state scale	Author reports ERFs of both response measures on LD for both B and A conditions After ERFs much lower, but somewhat parallel to the Before ERFs Shows strong excess response, equivalent to that of a 6 dB (8 dB for the second response measure) change in exposure No statistical test reported	This study has been included under Type A source interventions even though it is not fully clear exactly what intervention(s) were responsible for the change in response

Abbreviations used in this and all subsequent tables: N: number of participants; B: Before-study; ERF: Exposure response function; n.a.: not applicable or not available; A: After-study; P(s): Participant(s); CI: confidence interval; ISO: ISO annoyance scales (ISO_TS_15666_2003); mos: months; Q: questionnaire; %HA: Percentage Highly Annoyed; s.d.: standard deviation; B/A: Before and After study; SE: standard error.