OPINIOÑ

A Social Security disability benefits claimant is no longer entitled to benefits when substantial evidence demonstrates (1) “there has been any medical improvement in the [claimant’s] impairment” and (2) the claimant “is now able to engage in substantial gainful activity.” 42 U-S.C. § 423(f)(1). To determine whether there has been medical improvement, an administrative law judge (ALJ) must “compare the current medical severity” of the claimant’s impairment to the medical severity of the impairment “at the time of the most recent favorable medical decision that [the claimant] w[as] disabled or continued to be disabled.” 20 C.F.R. § 404.1594 (b)(7). This appeal addresses the question of how an ALJ should determine whether medical improvement has occurred in so-called “closed period” cases — ie., where the ALJ finds in a single decision that the claimant was disabled for a closed period of time but has since medically improved.

We conclude that in closed period cases an ALJ should compare the medical evidence used to determine that the claimant was disabled with the medical evidence existing at the time of asserted medical improvement. Although the ALJ in this case made the appropriate comparison, substantial evidence does not support the ALJ’s finding of medical improvement. We therefore reverse the judgment and remand with instructions to remand this case to the ALJ to calculate an award of benefits.

I

This appeal concerns an ALJ’s award of social security disability benefits to Emily Attmore, who applied for disability benefits in October 2008. After the agency denied Attmore’s claim for benefits, an ALJ held a hearing in July 2010 and determined that, due to her bipolar disorder, Attmore was disabled as of April 15, 2007, but had medically improved to the point she was no longer disabled beginning on March 24, 2009. In making these findings, the ALJ considered the following medical evidence.

Attmore was last gainfully employed on April 13, 2007, when she quit her job after a series of breakdowns that prevented her from getting out of bed. In 2008, she was hospitalized three times because of her bipolar disorder. In May, she was hospitalized for two weeks after a suicide attempt during which she reported hearing voices and hallucinating. Although she “initially did well” after her release, she began to hallucinate and became restless after she reduced her medications or stopped taking them altogether. In August, she was hospitalized again for one week after experiencing hallucinations and “extreme paranoia.” In October, she experienced “situational problems related to [a] significant other,” again went off her medications and stopped sleeping, and was hospitalized a third time after she was found wandering around and responding to voices in her head. Upon her release one month later, Attmore began seeing psychiatrist Dr. Robert M. Wolf, as well as one of her former therapists.

Dr. Wolfs treatment notes report mixed progress over the next year. On November 10, 2008, for example, Attmore was “struggling” and “expressing some paranoid ideas,” but two weeks later said she was “doing well,” and Dr. Wolf reported her mental state was “fairly stable.” Similarly, on February 13, 2009, though Attmore seemed “quite stable,” she reported she had been “pretty depressed” for several weeks, exhibiting “social isolation, sleeping a lot.” By March 5, she said she was still “somewhat socially isolative,” but otherwise “feeling much better,” and Dr. Wolf noted she was “actually doing well.” On March 23 — the final day the ALJ found Attmore was disabled — she again said she was “feeling pretty well” with “no specific complaints,” but one week later she told another therapist she had been “staying in her apartment for [d]ays at a time, not leaving and not getting ‘dressed.” She experienced a gradual worsening of her symptoms from August through October as she increasingly struggled with depression.

Attmore’s progress in 2010 likewise was mixed. In February, she began working with a clinical social worker who agreed to drive her to appointments because she “d[id]n’t feel comfortable riding public transportation.” On February 26, Attmore told her new doctor she was “doing well” and spoke of starting school, but continued to complain of heightened anxiety when there were “other stresses in her life.” Although she “endorse[d] satisfaction with her life” in early March, by late March she said she was getting depressed, and had another episode during which she heard voices and slept for 19 hours. In April, her symptoms flared up two more times, causing increased depression and poor sleep for one week.

Based on this medical evidence, the ALJ determined Attmore was disabled between April 15, 2007 and March 23, 2009. At issue here is the ALJ’s medical improvement finding, which rested on two conclusions. The ALJ first detailed Dr. Wolfs treatment notes from March 23, 2009 and concluded Attmore had “benefited from mental health treatment and medication management and ha[d] experienced gradual improvement in her symptoms.” The ALJ then cited additional treatment notes and concluded Attmore had “shown improvement in the area of social functioning.” Based in part on the medical improvement finding, the ALJ awarded Att-more benefits only for the closed period from April 15, 2007 through March 23, 2009. After the Appeals Council denied review, Attmore filed a civil action seeking review of the ALJ’s decision, and the district court granted summary judgment to the Commissioner. We have jurisdiction over Attmore’s appeal under 28 U.S.C. § 1291 and now reverse.

II

We review the district court’s decision sustaining the ALJ’s denial of social security benefits de novo and can reverse only if the ALJ’s findings are based on legal error or are not supported by substantial evidence in the record. See Molina v. Astrue, 674 F.3d 1104, 1110 (9th Cir. 2012). Substantial evidence is “ ‘more than a mere scintilla,’ but may be less than a preponderance.” Id. at 1110-11 (quoting Valentine v. Comm’r Soc. Sec. Admin., 574 F.3d 685, 690 (9th Cir. 2009)). If the evidence is “susceptible to more than one rational interpretation,” we are required to affirm. Id. at 1111. We cannot affirm, however, “simply by isolating a specific quantum of supporting evidence,” but “must consider the record as a whole, weighing both evidence that supports and evidence that detracts from the [Commissioner’s] conclusion.” Tackett v. Apfel, 180 F.3d 1094, 1098 (9th Cir. 1999) (citation omitted).

Ill

Attmore raises two challenges to the ALJ’s determination that she medically improved as of March 24, 2009. First, she contends the ALJ erred in making only a general comparison of the medical evidence without reference to a specific point in time. In her -view, the ALJ was required to specifically identify the dates being compared, and the ALJ’s failure to do so was legal error. Second, Attmore argues that even if the ALJ’s comparison were proper, substantial evidence does not support a finding of medical improvement. We agree only with her second contention.

A

We begin with Attmore’s argument that the ALJ failed to make the appropriate comparison in determining she had medically improved. Medical improvement is defined as “any decrease in the medical severity” of a recipient’s impairment, 20 C.F.R. § 404.1594 (b)(1), and requires a “comparison of prior and current medical evidence which must show that there have been changes (improvement) in the symptoms, signs or laboratory findings associated with that impairment(s),” id. § 404.1594(c)(1). The Commissioner’s regulations prescribe a specific baseline for this comparison: an ALJ must “compare the ... medical severity” of the impairment(s) “present at the time of the most recent favorable medical decision” — that is, when the claimant was last found disabled or continued to be disabled — “to the medical severity of that impairment(s)”. at the time of the comparison. Id. § 404.1594(b)(7).

Making this comparison is straightforward in ordinary termination cases where the ALJ finds a claimant is disabled (or continues to be disabled) in one decision and, in a later decision, finds the claimant has medically improved. In those cases, the “most recent favorable medical decision” is an earlier decision, and the severity of the claimant’s impairment at the time of that decision provides the relevant baseline for comparison. Attmore’s case, however, is a so-called closed period case, meaning the ALJ found — in the same decision — that she had been disabled for a closed period of time and had since medically improved. Because the ALJ made its findings of disability and medical improvement in a single decision, there was no “most recent favorable medical decision” for comparison. The Commissioner’s regulations have no direct application in these circumstances and thus do not neatly prescribe an appropriate baseline.

We nevertheless agree with the parties that the thrust of the Commissioner’s regulations applies in closed period cases. Congress enacted the medical improvement standard as a safeguard against the arbitrary termination of benefits. See Pickett v. Bowen, 833 F.2d 288, 292 (11th Cir. 1987). As other circuits have explained, Congress intended this safeguard to apply to all claimants, including those who receive benefits in closed period cases. See Waters v. Barnhart, 276 F.3d 716, 719 (5th Cir. 2002) (holding the medical improvement standard applies to closed period cases); Shepherd v. Apfel, 184 F.3d 1196, 1200 (10th Cir. 1999) (same); Pickett, 833 F.2d at 291-92 (same); Chrupcala v. Heckler, 829 F.2d 1269, 1274 (3d Cir. 1987) (same); see also Jones v. Shalala, 10 F.3d 522, 523-24 (7th Cir. 1993) (applying the medical improvement standard to a closed period case). Accordingly, an ALJ should “engage[] in the same decision-making process” in closed period cases as in ordinary termination cases. Waters, 276 F.3d at 719 .

What, then, is the appropriate baseline for comparison in a closed period case? The regulations require a comparison of “prior and current medical evidence,” 20 C.F.R. § 404.1594 (c)(1), and define the “most recent favorable medical decision” as “the latest decision involving a consideration of the medical evidence and the issue of whether [the claimant] w[as] disabled,” id. § 404.1594(b)(7). In other words, the relevant baseline normally is the medical evidence underlying the ALJ’s most recent disability determination. For closed period cases, the analogous baseline is the medical evidence used to determine the claimant was disabled. We therefore hold that, in closed period cases, the ALJ should compare the medical evidence used to determine the claimant was disabled with the medical evidence existing at the time of possible medical improvement.

We are satisfied the ALJ made the appropriate comparison here. The ALJ made extensive findings that Attmore was disabled from April 15, 2007 through March 23, 2009. The ALJ then found medical improvement as of March 24, 2009, noting Attmore had “benefited from mental health treatment and medication management” and “experienced gradual improvement in her symptoms.” As illustration, the ALJ described in detail Att-more’s symptoms as of March 24 and pointed to evidence indicating she had “shown improvement in the area of social functioning.” Attmore faults the ALJ’s analysis for not specifically identifying the baseline for comparison. But the ALJ’s references to “improvement” implied a comparison to Attmore’s condition during the disability period, which the ALJ had just discussed. We can therefore draw the “specific and legitimate inference[]” that the ALJ compared the medical evidence from the date of possible improvement to the medical evidence used to determine that Attmore was disabled. Magallenes v. Bowen, 881 F.2d 747, 755 (9th Cir. 1989). There was no legal error.

B

Attmore next argues substantial evidence does not support the ALJ’s finding of medical improvement. We agree.

Some evidence does support the ALJ’s conclusion that Attmore had shown improvement in her symptoms and social functioning. Dr. Wolfs report from March 23, 2009&emdash;which the ALJ discussed in detail&emdash;assessed Attmore as “generally doing well at this time” and noted she had “no specific complaints.” Dr. Wolf also reported Attmore’s hygiene was appropriate, her thought processes were organized and spontaneous and there was no evidence of active psychosis. Other treatment notes likewise demonstrate that&emdash;as the ALJ observed&emdash;Attmore discussed the possibilities of vocational rehabilitation and returning to school and was improving her social relationships. Taken in isolation, the evidence on which the ALJ relied suggests some medical improvement.

The ALJ was required, however, to examine this evidence in the broader context of Attmore’s impairment. See Holohan v. Massanari, 246 F.3d 1195, 1205 (9th Cir. 2001) (“That a person who suffers from severe panic attacks, anxiety, and depression makes some improvement does not mean that the person’s impairments no longer seriously affect her_”). An ALJ cannot simply' “pick out a few isolated instances of improvement over a period of months or years” but must interpret “reports of ‘improvement’ ... with an understanding of the patient’s overall well-being and the nature of her symptoms.” Garrison v. Colvin, 759 F.3d 995, 1017 (9th Cir. 2014); see also Scott v. Astrue, 647 F.3d 734, 740 (7th Cir. 2011) (“[Ajlthough Scott had improved with treatment, she nevertheless continued to frequently experience bouts of crying and feelings of paranoia. The ALJ was not permitted to ‘cherry pick’ from those mixed results to support a denial of benefits.”). In short, the examples an ALJ chooses “must in fact constitute examples of a broader development.” Garrison, 759 F.3d at 1018 .

We conclude the examples on which the ALJ relied here were not in fact indicative of a “broader development” in two re spects. First, the improvement the ALJ highlighted was only temporary. It is the nature of bipolar disorder that symptoms wax and wane over time. With respect to such impairments, “[i]mprovement ... that is only temporary will not warrant a finding of medical improvement.” 20 C.F.R. § 404.1594 (c)(3)(iv). Although the ALJ pointed to isolated signs of improvement, the ALJ could not find medical improvement on that basis unless the ups and the downs of Attmore’s development showed sustained improvement. See id. § 404.1594(b)(1) (Example 2).

Attmore’s improvement was not sustained. On March 30, 2009 — -just one week after Dr. Wolf reported Attmore was “generally doing well” — another therapist reported that Attmore was “losing functioning ability” and said she “ha[d] been staying in her apartment for [d]ays at a time, not leaving and not getting dressed.” Although Attmore showed some signs of progress in the following months, she said she was “feeling increasingly depressed” in late April, and Dr. Wolf noted she had “significant depressive and anxious themes” in her speech. In July, she “could not sleep for about 3 days” after a flare-up of her symptoms. Attmore then struggled with depression for three months. As her depression worsened in August to October, she betrayed “anxious and depressive themes with some mild suicidal ideation” and suffered “daily intrusive suicidal thoughts.”

Attmore continued to exhibit only temporary improvement in 2010. For example, in March — at the time when she was discussing vocational rehabilitation and school — her symptoms again worsened. One week before the start of school, Att-more was “getting depressed,” sleeping 12-13 hours per night and growing anxious. The day before school began, Att-more reported experiencing “increased depression and hearing voices on one occasion,” and sleeping for 19 straight hours. Attmore did feel “a bit better” after deciding not to go to school, but over the next month she had two more flareups, including one causing a week of poor sleep. Given the frequency and persistent severity of these episodes, the evidence compels the conclusion that there was no “broader development” of sustained improvement underlying the ALJ’s examples. Garrison, 759 F.3d at 1018 .

Second, although the ALJ observed some “improvement in the area of social functioning,” that improvement was quite limited. Even during the disability period, Attmore had discussed going back to school. That she did so during the medical improvement period thus was not itself a sign of improvement — especially because in neither period did she follow through with her plans. Attmore also visited her father during the asserted medical improvement period, but that visit caused her depression to return along with a week-long flareup of her symptoms. She later stopped communicating with him entirely. Finally, despite improvement in other aspects of social functioning, Attmore’s paranoia persisted, making her unable to use public transportation. Throughout the medical improvement period, Attmore con tinued to rely on others in order to leave her house, including to run errands or attend doctor’s appointments. If she had to take the bus somewhere, she would ask a friend to go with her. Although she did take the bus alone one time, that experience resulted in yet another weeklong flare-up of symptoms after she felt harassed at the bus stop. This aspect of Attmore’s social functioning, in short, did not “in fact constitute examples of a broader development.” Garrison, 759 F.3d at 1018 (concluding there was no sustained medical improvement where “some symptoms came and went ..., some symptoms persisted nearly the whole period ..., and still other symptoms appear to have remained a constant source of impairment”).

We conclude the ALJ erroneously focused on only temporary periods and isolated aspects of Attmore’s improvement that were not representative of the continuing severity of her symptoms. Because Attmore’s improvement was not sustained and was considerably limited in scope, we hold substantial evidence does not support the ALJ’s finding of medical improvement.

CONCLUSION

The ALJ properly compared the medical evidence, but erred in determining that Attmore had medically improved as of March 24, 2009. We therefore reverse the judgment and remand with instructions to the district court to remand to the ALJ for the calculation and award of benefits.

REVERSED AND REMANDED.

. The point is that medical improvement determinations require a comparison of two distinct bodies of medical evidence. In closed period cases, the body of evidence used to determine the claimant was disabled may pertain to only the disability onset date, see, e.g., Shepherd, 184 F.3d at 1202 , or to the entire disability period, see Newbold v. Colvin, 718 F.3d 1257, 1260 (10th Cir. 2013). In the former scenario, courts have looked to medical evidence pertaining to the disability onset date as the appropriate baseline for comparison. See, e.g., Shepherd, 184 F.3d at 1202 (comparing the claimant’s symptoms from the disability onset date to those on the date of medical improvement). In the latter scenario, the Tenth Circuit has looked to medical evidence pertaining to the entire disability period as the appropriate baseline for comparison. SeeNewboid, 718 F.3d at 1264-65 . Either way, we understand these cases as applying the same rule we adopt here: an ALJ must compare the body of medical evidence as of the date of possible medical improvement to that used to determine the claimant was disabled. . "[T]aking public transportation” is one of the "[a]ctivities of daily living” the Commissioner considers in determining the severity of a claimant's impairment. 20 C.F.R. pt. 404, subpt. P, app. 1, § 12.00(C)(1). It also relates to a claimant’s social functioning. See id. § 12.00(C)(2) ("Social functioning includes the ability to get along with others, such as family members, friends, ... or bus drivers."). Attmore's distrust of strangers caused her to have panic attacks when she used public transportation. The ALJ thus properly considered this factor as part of her social functioning in determining the severity of her impairment during the disability period.