Table 3.

Results of Kruskal–Wallis test for assessing differences in prescribing rates according to the quality score of the studies and source of data.

	L-dopa prescribing ratea			COMT inhibitors prescribing rate			All DAs prescribing rate			MAO-B inhibitors prescribing rate			Amantadine prescribing rate			Anticholinergics prescribing rate
	Median	Range	P valuea	Median	Range	P value	Median	Range	P value	Median	Range	P value	Median	Range	P value	Median	Range	P value
Quality score
(1–3)	37.38	26–48.76	0.091	3.58	3.53–3.63	0.245	36.73	11.20–62.26	0.825	5.57	3.88–7.27	0.575	6.71	NA (one study only)	0.895	25.44	NA (one study only)	0.285
(4–6)	70.92	21–100		4.80	0.24–13.31		26.45	9.21–76.92		10.81	2.10–31		5.46	1.10–44.23		5.46	2.91–43
>6	87.17	51–98.50		6.80	3.10–10.10		28.75	7.63–75		10.50	1.67–21		5.14	0.80–22.10		5.14	3.81–31.40
Source of data
Insurance-claims, prescription registries, or drug sales databases	53.78	21–90	0.009 b	5.61	1–13.31	0.245	29	9.21–75	0.825	9.63	2.10–24.60	0.575	6.58	1.10–22.10	0.895	18.18	2.91–43	0.285
Medical charts and administrative databases	83.60	43.73–94.80		6	0.24–10.10		25	7.63–32.04		8.90	1.67–21		5.10	2–17.20		22.90	3.81–40.40
Patients' interviews, questionnaires, or surveys	90.40	78.84–100		4	3.10–5.80		43.70	18–76.92		11.08	2.30–31		5	0.80–44.23		19.80	8.50–30.76

^aTest based on Kruskal–Wallis statistic, significance level at P < 0.05. ^bPost hoc analysis: insurance-claims, prescription registries, or drug sales databases vs medical charts and administrative databases P=0.234. Insurance-claims, prescription registries, or drug sales databases vs patients' interviews, questionnaires, or surveys, P=0.011. Medical charts and administrative databases vs patients' interviews, questionnaires, or surveys P=0.582.